Research ArticleDeveloping a Contextually Personalized HybridRecommender System

Aysun Bozanta and Birgul Kutlu

Department of Management Information Systems Bogazici University Istanbul 34342 Turkey

Correspondence should be addressed to Aysun Bozanta aysunbozantabounedutr

Received 4 June 2018 Revised 5 September 2018 Accepted 30 September 2018 Published 23 October 2018

Academic Editor Ramon Aguero

Copyright copy 2018 Aysun Bozanta and Birgul Kutlu )is is an open access article distributed under the Creative CommonsAttribution License which permits unrestricted use distribution and reproduction in anymedium provided the original work isproperly cited

It is hard to choose places to go from an endless number of options for some specific circumstances Recommender systems aresupposed to help us deal with these issues and make decisions that are more appropriate )e aim of this study is to recommendnew venues to users according to their preferences For this purpose a hybrid recommendation model is proposed to integrateuser-based and item-based collaborative filtering content-based filtering together with contextual information in order to get ridof the disadvantages of each approach Besides that in which specific circumstances the user will like a specific venue is predictedfor each user-venue pair Moreover threshold values determining the userrsquos liking toward a venue are determined separately foreach user Results are evaluated with both offline experiments (precision recall F-1 score) and a user study Both the experimentalevaluation with a real-world dataset and a user study of the proposed system showed improvement upon the baseline approaches

1 Introduction

Social media platforms are very rich data resources forresearchers to mine and gain insight into user preferences)e increasing use of location-related technologies enablesthe development of location-based-services )ereforelocation-based social networks (LBSNs) which have becomethe host of new possibilities for user interaction haveemerged )ese systems which facilitate users to share theirvisits and explore other locations have accumulated hugeamount of data about users with extensive use over timeLocation Recommendation Systems (LRSs) have been de-veloped by discovering embedded information from thesedata to provide location suggestion for the users

)ere are three main recommendation techniqueswhich are also applied for location recommendationcontent-based filtering (CBF) collaborative filtering (CF)and hybrid recommendation Content-based filtering uti-lizes the information about an item itself for recommen-dations and tries to find the most similar item with the userrsquosprevious preferences Collaborative filtering recommends anitem according to the similarity between one userrsquos

preferences and the preferences of other individuals Hybridapproaches which are the composition of at least twoexisting approaches have recently been awarded for theirability to improve prediction Contextual information(weather time date etc) is more important in travel andtourism domains )erefore context-aware recommendersystems should be widely used for location recommendationrather than other types of recommendations (productmovie music etc) However most of the recommendationengines fail to consider contextual information for locationrecommendation

Personalization which should be handled from differentangles is another issue for recommendation systems )eeffect of each variable used in the recommendation may varyamong different users For instance two people may like thesame places but in different contextual circumstances)erefore it is important to consider the changing effects ofcontextual variables on different users

In this study a contextually personalized hybrid locationrecommender system is developed For this purpose usersrsquocheck-in history visited location properties (distance cat-egory popularity and price) and contextual data (weather

HindawiMobile Information SystemsVolume 2018 Article ID 3258916 13 pageshttpsdoiorg10115520183258916

season date and time of visits) were collected from TwitterFoursquare and Weather Underground A hybrid approach(user-based collaborative filtering item-based collaborativefiltering content-based filtering and context-aware rec-ommendation) was applied and results were evaluated withboth offline experiments (precision recall F-1 score) anda user study )is study is an expanded version of theprevious study [1] )e scientific value of this study can belisted as below

(i) )ree different types of variables (user-relatedvenue-related (content) and contextual) thathave not been used together in existing recom-mender systems were used in one algorithm todevelop a novel recommender system

(ii) Artificial neural network algorithm was applied todetermine the weight of each algorithm (user-basedcollaborative filtering item-based collaborativefiltering content-based filtering and context-awarerecommendation) that was used when developingthe hybrid recommendation system

(iii) )reshold values determining the userrsquos liking to-ward a venue were determined separately for eachuser

(iv) A contextually personalized recommendation wasgenerated by determining which contextual cir-cumstances were more appropriate for each specificuser-venue pair

(v) Data sparsity problem was alleviated(vi) Overspecialization was lessened(vii) Cold start problem was partially solved

2 Related Work

Developing a location recommender system is very attractivefor researchers because of its importance in both academiaand business )erefore although its history is based on lessthan a decade there are many studies on this subject

Content-based algorithms utilize the content informationof a location in order to handle data sparsity problem thatmayoccur in CF algorithms Table 1 presents mostly used contentinformation variables )e category variable specifies thecategory of a venue (restaurant shopping center theater etc))e distance variable specifies the distance from the user (GPSlocation center of visited venues etc) to the venue )e tagvariable specifies tags that are given by the users (can bevisited with friends romantic etc) Tips and comments arespecified by the user about the venue )e popularity variabledenotes the value of a location specified by ratings number ofvisits etc )e tags and tipscomments variables are used forsentiment analysis which is not in the context of this studyFor this reason only the category distance and popularityvariables were selected

Collaborative filtering algorithms can be categorized asmemory-based and model-based In addition memory-based CF is divided into two categories user-based CFwhich considers user similarity for recommendation [35ndash37] and item-based CF which considers the item similarity

for recommendation [38 39] Data mining techniques suchas neural networks [40] Naıve Bayesian modeling [13 23]association rule mining [41] and SVD [2] are used formodel-based CF

)e contextual approach emerged after traditional ap-proaches which simply focus on the past preferences ofcustomers Context represents a set of surrounding condi-tions of a user-item pair and affects the relation betweenthem A context-aware recommender system may considereither user context (income profession age current userlocation mood and status etc) or environmental context(current time weather traffic conditions events etc)[14 42] Contextual information is very crucial especiallyfor location recommendation )e decisions of the users forvenue visits are generally based on environmental factorsrather than on their decisions about other things (buyinga product listening to music etc) Even though contextualinformation is critical rather than additional for locationrecommendation systems it is not used in existing systemscommonly and effectively In the literature it is emphasizedthat context-based algorithms are demanding for effectiverecommender systems [43ndash46]

Contextual information for location recommendationcan be specified with many variables Mostly used variablesare presented in Table 2

As can be seen from Table 2 time and weather condition(eg sunny rainy and snowy) variables have been used forlocation recommendation more frequently than other var-iables)erefore time and weather conditions are selected asthe contextual variables for this study

Each filtering approach has different drawbacks Forexample the disadvantages of CF are the cold start problemdata sparsity and scalability [57] On the other hand in-formation need about an item and overspecialization are thedrawbacks of the CBF [58] Hybrid approaches combine atleast two of the existing approaches and aim to minimize orremove the drawbacks of existing approaches which mayoccur when they are used individually )erefore hybridsystems which are the combination of some of these ap-proaches can be the solution for a better recommendationsystem [43 44] Even though some hybrid systems arepresented in the literature there are still untouched pointsfor performance improvement of location recommendersystems )ere are studies considering different hybrid al-gorithms for location recommendation [2 7 19 4850 51 59] Seven different types of hybridization techniquesare mentioned in the literature namely weighted switchingmixed feature combination cascade feature augmentationand metalevel [60]

Table 1 Mostly used content-based variables for locationrecommendation

Content-based variables ReferencesDistance [2ndash12]Category [7 9 11 13ndash22]Tag [19 23ndash27]Tipscomments [19 28ndash31]Popularity [7 11 32ndash34]

2 Mobile Information Systems

)is study aims to develop a personalized hybrid rec-ommendation system using both user and location simi-larity location-related properties (distance categorypopularity and price classification) and varying effects ofcontextual data (weather season date and time of visits)among different users Weighted hybridization method isused to achieve better performance and to have drawbacks ofany individual recommendation system

3 Methodology

31 Data Collection )e aim of this study is to recommendnew venues to the users according to their preferences)erefore a location-based social network should be chosento collect the necessary data For this purpose two popularsocial networks Twitter and Foursquare were chosen Inorder to crawl usersrsquo check-in history Twitter is used sinceFoursquare does not allow direct streaming of user check-ins Foursquare was chosen to collect the characteristics ofvarious venues since it is one of the most popular location-based social networks and provides the characteristics ofvarious venues with its API

)e REST API of Twitter which is popularly used fordesigning web APIs to use pull strategy for data retrieval wasused for this study )e REST APIndashldquoGET searchtweetsrdquowhich returns a collection of relevant tweets matchinga specified query was used When a user who linked hisheraccount with Twitter check-ins using Swarm a related tweetincluding all check-in data appears on the Twitter timelineFirstly Twitter user idsrsquo of users who ldquochecked-inrdquo on theSwarm application and shared their check-ins over theTwitter application were collected for a two-month periodAfter that all geocoded tweet history of collected userswhich goes back to 2011 was retrieved and stored )eTwitter APIs (for Twitter API version 11) were used tocollect dataset by embedding them into the PHP code anddataset was stored in MySQL database

When a user checks in using Swarm and shares thischeck-in on Twitter the related tweet includes a URLstarting with ldquohttpswwwswarmappcomrdquo which con-tains a venue id at the end )ose venue ids were sent to theFoursquare API (httpsapifoursquarecomv2venuesVENUE_ID) and venue name category latitude longi-tude check-in count visitor count tip count and priceclassification of venues were collected as venue attributes

Weather history data are collected from the ldquoWeatherUndergroundrdquo website Each check-in date is matched with

the date in weather history for related weather condition(sunny rainy snowy etc)

32 Data Preprocessing Data preprocessing helps totransform raw data into an understandable format Real-world datasets are mostly incomplete inconsistent and lackcertain behaviors Data preprocessing is necessary for pre-paring these raw datasets for further processing Data re-duction which is one of the data preprocessing steps wasapplied to the raw dataset in order to obtain results that aremore accurate

)e raw dataset consisted of 6738 users 60202 venuesand 226227 visits Data reduction was performed accordingto the following criteria

(i) Only Istanbul check-ins were retrieved in order toincrease visit frequencies

(ii) )ere are various main categories of venues inFoursquare For this study ldquorestaurantrdquo was chosenas main category and all related subcategories ofrestaurant were used because of intensive check-infrequency in restaurants

(iii) Users who visited only one venue were extracted(iv) Venues which were visited by only one user were

extracted

After that 1101 users 711 venues and 4694 visitsremained in the dataset

)e terms used in this study can be found in Table 3

33 Rating Foursquare does not provide the direct ratingsfor venues from each individual user )erefore rating wascalculated from linear normalization of the frequencies ina range of 1 to 5 for each user-venue pair If a userrsquosmaximum and minimum number of visits are equal thenthe rating was determined as 1

Rating un vn( 1113857 freq un vn( 1113857minusminfreq un( 1113857

maxfreq un( 1113857minusminfreq un( 11138571113888 1113889lowast 41113888 1113889 + 1

(1)

34 Distance )e latitude and longitude values of venueswhich were collected from Foursquare were converted intothe x y z coordinates (Equations (2)ndash(4))

x cos(longitude)lowast cos(latitude) (2)

y sin(longitude)lowast cos(latitude) (3)

z sin(latitude) (4)

User centers were calculated by taking the weightedaverage of x y z coordinates of all visits for each user inorder to understand hisher active area Euclidean distancefrom each venue to the user center was calculated and namedas distance variable (Equation (5))

Table 2 Mostly used contextual variables

Context-related variables ReferencesTime [9ndash12 19 21 47ndash52]Weather conditions [48 51 53]Temperature [48]Trip type [54]Origin citydestination city [54]Speed and travel direction [55]Transportation type [56]

Mobile Information Systems 3

)is study aims to develop a personalized hybrid rec-ommendation system using both user and location simi-larity location-related properties (distance categorypopularity and price classification) and varying effects ofcontextual data (weather season date and time of visits)among different users Weighted hybridization method isused to achieve better performance and to have drawbacks ofany individual recommendation system

3 Methodology

31 Data Collection )e aim of this study is to recommendnew venues to the users according to their preferences)erefore a location-based social network should be chosento collect the necessary data For this purpose two popularsocial networks Twitter and Foursquare were chosen Inorder to crawl usersrsquo check-in history Twitter is used sinceFoursquare does not allow direct streaming of user check-ins Foursquare was chosen to collect the characteristics ofvarious venues since it is one of the most popular location-based social networks and provides the characteristics ofvarious venues with its API

)e REST API of Twitter which is popularly used fordesigning web APIs to use pull strategy for data retrieval wasused for this study )e REST APIndashldquoGET searchtweetsrdquowhich returns a collection of relevant tweets matchinga specified query was used When a user who linked hisheraccount with Twitter check-ins using Swarm a related tweetincluding all check-in data appears on the Twitter timelineFirstly Twitter user idsrsquo of users who ldquochecked-inrdquo on theSwarm application and shared their check-ins over theTwitter application were collected for a two-month periodAfter that all geocoded tweet history of collected userswhich goes back to 2011 was retrieved and stored )eTwitter APIs (for Twitter API version 11) were used tocollect dataset by embedding them into the PHP code anddataset was stored in MySQL database

When a user checks in using Swarm and shares thischeck-in on Twitter the related tweet includes a URLstarting with ldquohttpswwwswarmappcomrdquo which con-tains a venue id at the end )ose venue ids were sent to theFoursquare API (httpsapifoursquarecomv2venuesVENUE_ID) and venue name category latitude longi-tude check-in count visitor count tip count and priceclassification of venues were collected as venue attributes

Weather history data are collected from the ldquoWeatherUndergroundrdquo website Each check-in date is matched with

the date in weather history for related weather condition(sunny rainy snowy etc)

32 Data Preprocessing Data preprocessing helps totransform raw data into an understandable format Real-world datasets are mostly incomplete inconsistent and lackcertain behaviors Data preprocessing is necessary for pre-paring these raw datasets for further processing Data re-duction which is one of the data preprocessing steps wasapplied to the raw dataset in order to obtain results that aremore accurate

)e raw dataset consisted of 6738 users 60202 venuesand 226227 visits Data reduction was performed accordingto the following criteria

(i) Only Istanbul check-ins were retrieved in order toincrease visit frequencies

(ii) )ere are various main categories of venues inFoursquare For this study ldquorestaurantrdquo was chosenas main category and all related subcategories ofrestaurant were used because of intensive check-infrequency in restaurants

(iii) Users who visited only one venue were extracted(iv) Venues which were visited by only one user were

extracted

After that 1101 users 711 venues and 4694 visitsremained in the dataset

)e terms used in this study can be found in Table 3

33 Rating Foursquare does not provide the direct ratingsfor venues from each individual user )erefore rating wascalculated from linear normalization of the frequencies ina range of 1 to 5 for each user-venue pair If a userrsquosmaximum and minimum number of visits are equal thenthe rating was determined as 1

Rating un vn( 1113857 freq un vn( 1113857minusminfreq un( 1113857

maxfreq un( 1113857minusminfreq un( 11138571113888 1113889lowast 41113888 1113889 + 1

(1)

34 Distance )e latitude and longitude values of venueswhich were collected from Foursquare were converted intothe x y z coordinates (Equations (2)ndash(4))

x cos(longitude)lowast cos(latitude) (2)

y sin(longitude)lowast cos(latitude) (3)

z sin(latitude) (4)

User centers were calculated by taking the weightedaverage of x y z coordinates of all visits for each user inorder to understand hisher active area Euclidean distancefrom each venue to the user center was calculated and namedas distance variable (Equation (5))

Table 2 Mostly used contextual variables

Context-related variables ReferencesTime [9ndash12 19 21 47ndash52]Weather conditions [48 51 53]Temperature [48]Trip type [54]Origin citydestination city [54]Speed and travel direction [55]Transportation type [56]

Mobile Information Systems 3

distance

(xminus center x)2 +(yminus centery)2 +(zminus center z)21113969

(5)

35 Popularity Foursquare provides four variables abouta venue check-in count like count user count and tipcount In this study these variables were considered asreference to the popularity of a venue )erefore thepopularity variable was formed from these four properties ofvenues by applying Principal Component Analysis (PCA)which aims at dimension reduction Before applying PCAsampling adequacy for PCA should be checked For thispurpose KMO and Barlettrsquos tests were used Samplingadequacy can be observed in Table 4 which presents KMOvalue as 0821 and significance of Barlettrsquos test of sphericityas 0001 )e acceptable level of KMO is generally 06 andBarlettrsquos test of sphericity is significant at the 1 alpha level)e results showed that the sample is adequate for PCA

Ninety-three percent of the total variance was explainedby only one component (Table 5) )erefore it can beconcluded that one variable which was named ldquopopularityrdquocan be used instead of four variables

)e component matrix shows the correlation betweenvariables and component Since the correlation values rangefrom minus1 to +1 it can be concluded that there is a strongpositive correlation between a component and each of thevariables (Table 6)

36 Category All subcategories of food which were col-lected by the FS API were included in this study )ere are34 restaurant categories including different countriesrsquo cui-sine in the dataset User-category matrix which presents the

number of visits of each user in each subcategory wasprepared

37 Price )ere is a four-price classification in Foursquare1-cheap 2-average 3-expensive and 4-very expensive )euser-price matrix presenting the number of visits of eachuser in each price class was prepared by using the datacoming from FS API

38 Time Twitter provides UNIX time format for each tweetin order to understand date and time it is converted to dateand time stamp For this study season day and the differentperiods of the day were used as contextual variables It wasobserved that some of the values of some contextual variablesshowed similar characteristics such as users have the samepattern of check-in behaviors for weekdays )erefore dis-cretization was applied to the contextual variables whichdisplayed a better performance Days were discretized asldquoweekdayrdquo and ldquoweekendrdquo [51 61] )e check-ins that weremade in spring or summer were categorized as ldquohot seasonrdquocheck-ins while the check-ins that were made in autumn andwinter were categorized as ldquocold seasonrdquo check-ins [61] In thestudies of Majid et al [48] and Wang et al [62] time isdiscretized as morning afternoon evening and night In thestudy [61] a day is discretized as morning and evening onlyHowever after exploring the check-in behaviors in thedataset it is found that discretization as morning noon andevening would bemore suitable)e time range 0700 to 1159was defined as ldquomorningrdquo 1200 to 1659 as ldquonoonrdquo and 1700to 0659 as ldquoeveningrdquo )us it is aimed to make a moreaccurate recommendation

39 Weather )e data of weather condition which is alsoa contextual variable were collected from the WeatherUnderground API that provides more than 10 differentweather conditions (sunny rainy snowy rainy and stormy

Table 3 Terms used for the study

Terms Definition

freq(un vn)Number of visits from nth user to the nth

venueminfreq(un) Minimum number of visits of nth usermaxfreq(un) Maximum number of visits of nth useruser_sim( u

rarrn u

rarrm) Similarity between nth and mth users

urarr

n Vector consisting of ratings of nth userurarr

m Vector consisting of ratings of mth userRating(un vn) Rating of nth user to nth venuevenue_sim ( v

rarrn v

rarrm) Similarity between nth and mth venues

vrarr

n Vector consisting of ratings of nth venuevrarr

m Vector consisting of ratings of mth venue

Check-in count Total number of check-ins in a specificvenue

Like count Total number of people who liked thespecific venue

User count Total number of unique people who checkin specific venue

Tip count Total number of comments for specificvenue

Rating

Score of each venue which is provided byFS API and calculated from the scoring ofall people who check in that specific venue

out of 10

Table 4 KMO and Barlettrsquos test results

KMO measure of sampling adequacy 0821

Barlettrsquos test of sphericityApproximate chi-square 13951963Degrees of freedom 6

Significance 0001

Table 5 Total variance explained

Component of variance Cumulative variance1 9369 93692 446 98153 108 99194 080 100

Table 6 Component matrix for popularityCheck-in count 0965Like count 0985User count 0981Tip count 0941

4 Mobile Information Systems

snowy and stormy etc) It was observed that some weathercategories showed the same patterns of user behavior)erefore they were discretized under three main cate-gories ldquosunnyrdquo ldquorainyrdquo (all categories including rainy) andldquosnowyrdquo (all categories including snowy) )e percentage ofwhich contextual circumstances the venue is preferred wascalculated by the proportion of visits in the specific categoryover the total visits

310 Development of Recommendation System Developmentsteps of this recommendation system are explained in detailbelow

User similarity values were calculated from the user-venue matrix which presents the ratings of users to thevenues with cosine similarity (Equation (6)) Ratings(ratingucf ) of the user to the other venues were predicted byusing user similarity values (Equation (7))

user_sim urarr

n urarr

m( 1113857 cos urarr

n urarr

m( 1113857 urarr

n urarr

m

urarr

n

u

rarrm

(6)

ratingucf 1113936 user_sim un um( 1113857 times rating um vn( 1113857

1113936 user_sim un um( 1113857 (7)

User similarity values were calculated from the user-category matrix (Table 7) which presents the number ofvenues that a specific user visited for each category withCosine similarity in a similar manner as in Equation (2)

Popularity values were discretized into three cate-gories namely high medium and low according totheir normalized popularity values which were attainedfrom PCA )en the user-popularity preference matrix(Table 8) was generated which keeps the number ofvenues that a specific user visited in each popularitycategory

User similarity values were calculated from the user-popularity matrix which presents the number of venuesthat a specific user visited in each popularity class with

cosine similarity in the similar manner as in Equation(2)

)e user-price preference matrix (Table 9) was alsoconstructed which keeps the number of venues that a spe-cific user visited in each price category User similarityaccording to price preferences was also calculated in thesimilar manner as in Equation (2)

Equation (3) was also used to calculate predicted ratingsusing user similarity which depend on the category(ratingcategory) popularity (ratingpopularity) and price(ratingprice) preferences of users

Venue similarity values were calculated from the user-venue matrix with cosine distance (Equation (8)) Ratings(ratingicf ) of the user to the other venues were also predictedby using venue similarity values (Equation (9))

venue_sim vrarr

n vrarr

m( 1113857 cos vrarr

n vrarr

m( 1113857 vrarr

n vrarr

m

vrarr

n

v

rarrm

(8)

ratingicf 1113936 venue_sim vn vm( 1113857 times rating un vm( 1113857

1113936 venue_sim vn vm( 1113857

(9)

)e venue-context matrix (Table 10) which keeps trackof the contextual features of the venues was prepared )epercentages showing the venue preferences in differentcontextual circumstances were presented in this matrixWith the help of this matrix venue similarities were alsocalculated using Equation (4) Predicted ratings(ratingcontext) are also calculated as in Equation (5) using thecontextual similarity of venues

)e calculation of ratings (ratingdistance) according to thedistance between a venue to the user depends on the as-sumption that if this distance is short then the user will visitthat venue more frequently [63ndash66] Although the users are

more willing to check in at nearby venues to their centersdistance perception of each user is different Optimal co-efficients (A B and n) for power law distribution [49 64 65]were determined to model the willingness of a user to go andcheck in at a place

ratingdistance A + Blowast distancen (10)

In this study a weighted hybridization technique wasused to compute the score of recommended items using allavailable recommendation algorithms Artificial neuralnetwork (ANN) analysis was applied in order to find theoptimal weights for each technique instead of using equalweights for them Inputs to the ANN were the results of

Mobile Information Systems 5

all available recommendation algorithms and the outputto be predicted was the actual rating Final ratings

were calculated by multiplying the ANN weights andratings

Rating w1 lowast ratingucf + w2 lowast ratingicf + w3 lowast ratingdistance+ w4 lowast ratingpopularity + w5 lowast ratingcategory+ w6 lowast ratingprice + w7 lowast ratingcontext

(11)

In order to decide whether to recommend a venue to theuser or not different threshold values were used for eachuser )e threshold value for each user was determined bytaking the average ratings of that user After that if thecalculated rating (Equation (11)) was greater than hisherthreshold then it was considered that the user will like thatvenue )is is the first version of the algorithm and it isnamed as ldquoHybRecSysrdquo

Existing recommender systems do not consider that thepreferences of the users are affected by different contextualcircumstances For instance a user may prefer a venue ona rainy weekday at noon while another user may prefer thesame venue in another context In order to handle this issueour system calculates the probability of visiting a venue ina specific contextual category For instance the followingequation calculates the probability of visit of user i to thevenue j in the mornings

P time morning visitij111386811138681113868111386811138681113874 1113875

1113936kcontextual similarityjk lowast probablity of morningik

1113936kcontextual similarityjk

(12)

For each user-venue pair there are 36 different con-textual circumstances (day weekday weekend time

morning noon evening season hot cold weather

sunny rainy snowy) For each situation probabilities werecalculated and the resulting table was constructed(Table 11)

Table 11 respectively presents user id venue id averagerating of related user the percentage that the user will visitthat venue in that time category the percentage that the userwill visit that venue in that day category the percentage thatthe user will visit that venue in that season category thepercentage that the user will visit that venue in that weathercategory total point from all contextual variables (sum of allpercentages) predicted rating and the final decision (Like)

)e sum of all categories of each contextual variableshould add up to one For instance since the day variable hastwo categories weekdays and weekend if a userrsquos probabilityof visiting a specific venue on a weekday is 06 then theprobability that user will visit the same venue on a weekendhas to be 04 )erefore the sum of the values of all con-textual variables (Contexttotal) may have a maximum valueof 4 )e final decision of whether a user will like a venue ornot depends on two things the predicted rating havinga greater value than the average rating of the user and thetotal context having a value of at least 2 out of 4 )e finalversion of the algorithm was called ldquocontextually

personalized HybRecSysrdquo and the whole developmentprocess is depicted in Figure 1

4 Evaluation of Recommendation System

)e performance evaluation of the proposed system will beexplained in detail in this section For the evaluation ofrecommender systems there are three types of experimentalsettings [67]

(1) Offline experiments in which a precollected datasetof users is used to validate the results

(2) User studies in which a small sample of users areasked to perform several tasks requiring an in-teraction with the recommendation system

(3) Online evaluation trying to evaluate different algo-rithms over online recommender systemwith a smallpercentage of the traffic

41 Offline Experiments )e Contextually personalizedHybRecSys was compared with five algorithms user-basedK-nearest neighborhood (KNN) [68] item-based KNN [69]biased matrix factorization [70] SVD++ [71] andHybRecSys [1] First four algorithms were available in theLibRec a Java library for recommender systems For each

Table 7 Sample of user-category matrix

User id Turkish_rest American_rest Chinese_rest11949 10 0 010147822 0 6 210437542 0 2 0

Table 8 Sample of user-popularity matrix

User id Low_popularity Medium_popularity High_popularity

11949 7 3 010147822 0 5 310437542 2 0 0

Table 9 Sample of user-price matrix

User id 1 2 3 411949 6 3 1 010147822 0 0 3 510437542 2 0 0 0

6 Mobile Information Systems

algorithm the default settings of LibRec were used Fifthalgorithm HybRecSys is the earlier version of ContextuallyPersonalized HybRecSys Other hybrid algorithms definedin the literature could not be included in this study since thesource codes were not available

Precision recall and F-1 measures were used as eval-uation metrics in this study Precision specifies the per-centage of correctly recommended items over totalrecommended items Recall indicates the percentage ofrecommended items over the total number of liked items bythe user)e F1 scoremeasures the accuracy of the system byusing both precision and recall

In order to split the data into training and test sets K-fold (K 10) cross-validation technique was used )edataset was divided into 10 disjoint sets making sure that

each set contains about 10 of the visits of each user One setwas used as the test set and nine sets were used as the trainingset for each fold

Figure 2 shows precision recall and F1 measure of eachalgorithm and it is obvious that Contextually PersonalizedHybRecSys outperforms all other algorithms

According to precision metric HybRecSys user-basedKNN item-based KNN biased matrix factorization andSVD++ follow contextually personalized HybRecSys re-spectively According to recall metric HybRecSys biasedmatrix factorization item-based KNN SVD++ and user-based KNN follow contextually personalized HybRecSysAccording to F-1 measure HybRecSys item-based KNNbiased matrix factorization user-based KNN and SVD++follow contextually personalized HybRecSys respectively

User-venuematrix

User-categorypreferences

matrix

User-popularitypreferences

matrix

User-pricepreferences

matrix

Venue-contextual

feature matrix

UsersVenues

Check-in historyVenue properties

(category popularity price)Contextual data

(weather time frame days season)

Distance

Contextualprobabilities

Ratingcontext

Contexttotal Rating Average rating of a user Recommend2 amp

Ratingprice Ratingpopularity Ratingcategory RatingdistanceRatingicf Ratingucf

Figure 1 Framework of contextually personalized HybRecSys

Table 11 )e final decision table

Uid Vid Av_rating Time Day Season Weather Contexttotal Rating Likeu1 v1 25 Morning 05 Wdays 0 Hot 1 Sunny 1 25 3 Trueu1 v1 25 Morning 05 Wdays 0 Hot 1 Rainy 0 15 3 Falseu1 v1 25 Morning 05 Wdays 0 Hot 1 Snowy 0 15 3 Falseu1 v1 25 Morning 05 Wdays 0 Cold 0 Sunny 1 15 3 False

Table 10 Sample of venue-context matrix

Venue id Hot_season Cold_season W_day W_end Morning Noon Evening Sunny Rainy SnowyVenue1 075 025 068 032 008 017 075 067 025 008Venue2 05 05 05 05 0 1 0 1 0 0Venue3 083 017 096 004 013 052 035 065 035 0

Mobile Information Systems 7

42 User Study for Contextually Personalized HybRecSysAs was indicated in the literature user studies are veryhelpful to understand whether the recommendations areliked by the users and to collect more detailed data about therecommendation system [45 57 67] Although conductinga user study is difficult time-consuming and costly it issuggested to apply them after the offline experiments inorder to validate the results of the offline experiments [67]

43 Steps of the User Study For this study a user study wasconducted on the users in our dataset For this purpose theTwitter account of each user in our dataset was checked tolearn whether their profiles allow to receive direct messagesOut of 1101 users 195 accounts were open for directmessage )ose users were invited to attend our user studyby direct message and a small incentive (a movie ticket) waspromised if they attend this evaluation

Twenty-four users replied to themessage and accepted toattend the study After that the user evaluation occurred inthe following steps

(1) )e algorithm predicted the ratings of that user to allvenues except the venues they visited

(2) Among the results the algorithm recommendedthree top-rated venues for each 24 users )ese userswere asked via Twitter message whether any of therecommended items attracted their attention

(3) Twenty-one users replied to the recommendationsOnly one of them said that none of the venues wassuitable for them Others were interested in at leastone venue among the recommended ones

(4) A survey which also includes the Foursquare links ofrecommended venues was prepared according to theparticipantsrsquo choices and sent to them

(5) )e users were asked to fill out the survey after theyvisit that recommended venue or after examiningthe Foursquare page of the venues

44 Survey Questions of the User Study )ere are 12 ques-tions in the survey )e first question was asked to un-derstand the appreciation of the participants to the

recommended venues It was asked in 5-point Likert scale (1-Not Like At All 2-Not Like 3-Not Sure 4-Like 5-Like VeryMuch)

)e following four questions were asked to measure theappropriateness of the category price popularity and thelocation of the recommended venue for the participant)eywere also asked using a 5-point Likert scale (1-Not Ap-propriate At All 2-Not Appropriate 3-Not Sure 4-Ap-propriate 5-Very Appropriate)

)e following four questions were asked to understandin which contextual circumstances the user would prefer therecommended venue )ese questions were asked as fixedsum scale questions )e participants were asked to dis-tribute a hundred points to the categories of a contextualvariable according to the tendency of the user to visit thatvenue in these categories

)e last two questions were demographic questions)ey were asked to learn the age gender and the educationlevel of the participants

45 Results of the User Study Participantsrsquo average age is 30and the age range varies from 19 to 38 )ere are eightwomen and 12 men in the dataset 12 of the participantsgraduated from high school six of them have a bachelorrsquosdegree and two of them have a masterrsquos degree Table 12presents the answers of the participants to the first fivequestions of the survey

Ninety percent of the participants liked the recom-mended venues Ten percent of them were indecisive aboutwhether they like the recommended venue Ninety percentof the participants thought that the price class of the rec-ommended venue was suitable for them Forty percent of theparticipants thought that the category of recommendedvenues were very appropriate for them while 20 of themthought that the categories were appropriate and theremaining are indecisive Seventy percent of the participantsthought that the popularity class of recommended venueswas appropriate for them and 10 of them thought that thepopularity classes of the venues were very appropriate Onthe other hand 10 were indecisive while another 10thought that the popularity classes of the venues were notappropriate Moreover the address of the recommended

05

04

03

02

01

0User-based

KNNItem-based

KNNBiased matrixfactorization

SVD++ HybRecSys(neural)

ContextualpersonalizedHybRecSys

PrecisionRecallF1-measure

Figure 2 Graph of precision recall and F-1 measures of the algorithms

8 Mobile Information Systems

venue was given to the participants and they were askedwhether the location was appropriate or not 40 of theparticipants said very appropriate and 30 of them saidappropriate while the other 30 said not appropriate

)e following four questions were asked to understandin which contextual circumstances participants will prefer tovisit the recommended venues Table 13 presents the an-swers of each participant to these questions and the pre-dictions which are calculated from the algorithms for eachuser

)e predicted results were compared with the actualanswers of the participants )e evaluation of the recom-mendation was measured with precision recall and F-1measures Table 14 presents the precision recall and F-1scores for 20 participants Precision value is 0282 recallvalue is 0276 and F1-score is 0279

)e participants made some comments about the venuesvia Twitter messages Some of the participants had been tosome of the restaurants and supported the given recom-mendation with the following expressions

Table 12 Survey answers (Q1ndashQ5)

1 () 2 () 3 () 4 () 5 ()1 Rate your liking 10 80 102 Rate the appropriateness of the price range 10 903 Rate the appropriateness of the category of therestaurant 40 20 40

4 Rate the appropriateness of the popularity class ofthe restaurant 10 10 70 10

5 Rate the appropriateness of the location of therestaurant 30 30 40

Table 13 Answers of questions 6-7-8-9 and prediction of the answers

Answers of questions 6-7-8-9 Prediction of the answersM N E Wdays Wend Hot Cold Sunny Rainy Snowy M N E Wdays Wend Hot Cold Sunny Rainy Snowy

1 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 0002 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 0003 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 0004 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 0005 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 0006 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 0007 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 0008 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 0009 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02510 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 00011 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 00012 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 00013 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 00014 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 00015 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 00016 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 00017 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 00018 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 00019 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02520 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 000

Table 14 Precision recall and F-1 measure for user study

Approaches Precision Recall F-1 scoreUser-based KNN 01220 00823 00983Item-based KNN 01154 01107 01130Biased matrix factorization 00893 01200 01031SVD++ 00702 00976 00816HybRecSys 01667 01460 01493Contextually personalized HybRecSys 018 045 025User study results of contextually personalizedHybRecSys 0282 0276 0279

Mobile Information Systems 9

(i) Participant 1 (Gender Male Age 37 EducationHigh School)

(a) ldquoI always prefer going these two restaurants thatyou have recommendedrdquo

(ii) Participant 2 (Gender Male Age 38 EducationHigh School)

(a) ldquoI have already been to one of the restaurantsrdquo

(iii) Participant 3 (Gender Female Age 19 EducationHigh School)

(a) ldquoI have visited one of the restaurants beforerdquo

(iv) Participant 4 (Gender Female Age 38 EducationBachelorrsquos Degree)

(a) ldquoI have already visited venue1 and venue2rdquo

)ese statements demonstrate the ability of the rec-ommender systemrsquos accurate prediction Even participantsthat had not visited the recommended venues before statedthat they like the recommended venues )is result revealsthat the developed system has the diversity novelty andserendipity dimensions

5 Conclusion

In this study a contextually personalized hybrid recom-mendation model was proposed)is model integrates user-based and item-based collaborative filtering content-basedfiltering together with contextual information in order to getrid of the drawbacks of each approach Different datasources were used to collect the data Visiting history of userswas collected from Twitter Venue characteristics (distancecategory popularity and price classification) were collectedfrom Foursquare and contextual information (weatherseason date and time of visits) related to each visit werecollected from Weather Underground website For content-based filtering the variables distance category popularityand price classification that have not been used before in onealgorithm were used in order to determine content-baseduser similarity Weather conditions season date and timeof each visit were used cumulatively as the properties ofvenues and contextual similarities of venues were utilized)e artificial neural network algorithm was applied to de-termine the weights of each algorithm Ratings coming fromdifferent algorithms (user-based CF item-based CFcontent-based filtering and rating calculated from thecontextual similarities of venues) were used as the predictorsof the actual rating Final ratings were calculated by mul-tiplying the weights retrieved from neural network and theratings from different algorithms In addition in order tomake a more accurate recommendation and make therecommender system contextually personalized our systemcalculates the probability of visiting a venue in a specificcontextual category for each user-venue pair)e decision ofrecommendation of a venue to the specific user is madeaccording to two rules If the calculated rating is greater thanthe average rating of the user and the total contextual score isgreater than two then that venue will be recommended to

that user under the specific contextual circumstances )eHybrid system prunes the disadvantages of each approachthat may occur when they are used separately

Contextually Personalized HybRecSys was compared touser-based and item-based KNN biased matrix factorizationand SVD++ )ese algorithms are evaluated according to themetrics which are used for ranking prediction (precisionrecall and F-1 measure) Training and test datasets werecreated using K-fold cross-validation (K 10) techniqueResults show that contextually personalized HybRecSysoutperforms existing four algorithms according to eachevaluation metric Contextually Personalized HybRecSys ef-fectively overcomes the data sparsity problem using venuecategory popularity and price to model user preferences

In addition recommending very similar venues toprevious visits causes overspecialization )is problem isalso reduced by considering the user preferences with dif-ferent aspects and not just being stuck on only the venuecharacteristics Hence the quality of the recommendation isimproved and the recommender system gained diversitynovelty and serendipity dimensions

)e algorithm partially solves the cold start issue whichcan be caused by both a new user and a new item Even ifa new user rates only one venue the algorithm understandsthe user preferences from the characteristics of the venue(category price and popularity) Moreover the algorithmfigures out the contextual circumstances under which theuser prefers that the specific venue )erefore by looking atthese characteristics the algorithmmay recommend a venueto a new user

)e most important feature that distinguishes the de-veloped algorithm from others is the contextual personal-ization Contextually personalized HybRecSys solvesthis issue and recommends a right venue under the rightconditions

At the beginning the collected data size was large butafter the filtration it became moderately small On theother hand this small dataset generated significant resultsthat may be improved with larger datasets As a futurestudy it is planned to apply contextually personalizedHybRecSys on larger datasets Venue opening and closinghours can be checked for better results In order to solvethe cold start problem entirely the system may recom-mend a venue to a new user by looking at the contextualcircumstances and recommend the most preferred venuebased on these contextual circumstances In additioneven if a new venue is added to the system it can berecommended by looking at the content-related charac-teristics )e online evaluation was conducted with only20 people which is a relatively small sample for obtainingstatistically precise and coherent results )erefore moreusers should be reached for future studies Finally thisalgorithm can be embedded into a mobile application andthis mobile application can be marketed in mobile ap-plication stores )erefore its performance can be mea-sured via online experiments As users download and usethe application they will score the recommendation sothat online feedback can be taken and the algorithm can beimproved

10 Mobile Information Systems

Data Availability

)e Twitter and Foursquare data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

)e authors declare that they have no conflicts of interest

Acknowledgments

)is study was funded by Bogazici University ScientificResearch Fund (grant no 11463)

References

[1] A Bozanta and B Kutlu ldquoHybRecSys content-based con-textual hybrid venue recommender systemrdquo Journal of In-formation Science pp 1ndash15 2018

[2] M Sattari I H Toroslu P Karagoz P Symeonidis andY Manolopoulos ldquoExtended feature combination model forrecommendations in location-based mobile servicesrdquoKnowledge and Information Systems vol 44 no 3 pp 629ndash661 2015

[3] H Wang G Li and J Feng ldquoGroup-based personalizedlocation recommendation on social networksrdquo in WebTechnologies and Applications pp 68ndash80 Springer BerlinGermany 2014

[4] H Yin B Cui Y Sun Z Hu and L Chen ldquoLCARS a spatialitem recommender systemrdquo ACM Transactions on In-formation Systems (TOIS) vol 32 no 3 pp 1ndash37 2014

[5] M H Park J H Hong and S B Cho ldquoLocation-basedrecommendation system using bayesian userrsquos preferencemodel in mobile devicesrdquo in Ubiquitous Intelligence andComputing pp 1130ndash1139 Springer Berlin HeidelbergGermany 2007

[6] K Aihara H Koshiba and H Takeda ldquoBehavioral cost-basedrecommendation model for wanderers in townrdquo in Human-Computer Interaction Towards Mobile and IntelligentInteraction Environments pp 271ndash279 Springer BerlinHeidelberg Germany 2011

[7] Z Yu Y Feng H Xu and X Zhou ldquoRecommending travelpackages based on mobile crowdsourced datardquo IEEE Com-munications Magazine vol 52 no 8 pp 56ndash62 2014

[8] H Yin B Cui L Chen Z Hu and C Zhang ldquoModelinglocation-based user rating profiles for personalized recom-mendationrdquo ACM Transactions on Knowledge Discovery fromData vol 9 no 3 pp 1ndash41 2015

[9] D Zhou S M Rahimi and X Wang ldquoSimilarity-basedprobabilistic category-based location recommendation uti-lizing temporal and geographical influencerdquo InternationalJournal of Data Science and Analytics vol 1 no 2 pp 111ndash121 2016

[10] H Zhu E Chen H Xiong K Yu H Cao and J TianldquoMining mobile user preferences for personalized context-aware recommendationrdquo ACM Transactions on IntelligentSystems and Technology vol 5 no 4 pp 1ndash27 2015

[11] Q Fang C Xu M S Hossain and M G Stcaplrs ldquoA spatial-temporal context-aware personalized location recommen-dation systemrdquo ACM Transactions on Intelligent systems andTechnology vol 7 no 4 pp 1ndash30 2016

[12] Q Yuan G Cong K Zhao Z Ma and A Sun ldquoWho wherewhen and what a nonparametric bayesian approach to

context-aware recommendation and search for twitter usersrdquoACM Transactions on Information Systems vol 33 no 1pp 1ndash31 2015

[13] A Gupta and K Singh ldquoLocation based personalized res-taurant recommendation system for mobile environmentsrdquoin Proceedings of International Conference on Advances inComputing Communications and Informatics (ICACCI)pp 507ndash511 IEEE Mysore India August 2013

[14] J Bao Y Zheng and M F Mokbel ldquoLocation-based andpreference-aware recommendation using sparse geo-socialnetworking datardquo in Proceedings of 20th International Con-ference on Advances in Geographic Information Systemspp 199ndash208 Redondo Beach CA USA November 2012

[15] H Yin Y Sun B Cui Z Hu and L Chen ldquoLcars a location-content-aware recommender systemrdquo in Proceedings of 19thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 221ndash229 Chicago IL USAAugust 2013

[16] M H Kuo L C Chen and C W Liang ldquoBuilding andevaluating a location-based service recommendation systemwith a preference adjustment mechanismrdquo Expert Systemswith Applications vol 36 no 2 pp 3543ndash3554 2009

[17] K Shimada H Uehara and T Endo ldquoA comparative study ofpotential-of-interest days on a sightseeing spot recom-menderrdquo in Proceedings of IIAI 3rd International Conferenceon Advanced Applied Informatics (IIAIAAI) pp 555ndash560Kitakyushu Japan August 2014

[18] L Ahmedi K Rrmoku K Sylejmani and D Shabani ldquoAbimodal social network analysis to recommend points ofinterest to touristsrdquo Social Network Analysis and Miningvol 7 no 1 p 14 2017

[19] H Gao J Tang X Hu and H Liu ldquoContent-aware point ofinterest recommendation on location-based social networksrdquoin Proceedings of the AAAI pp 1721ndash1727 Austin TexasUSA January 2015

[20] Z Zheng Y Chen S Chen L Sun and D Chen ldquoLocation-aware POI recommendation for indoor space by exploitingWiFi logsrdquo Mobile Information Systems vol 2017 Article ID9601404 16 pages 2017

[21] W X Zhao N Zhou A Sun J R Wen J Han andE Y Chang ldquoA time-aware trajectory embedding model fornext-location recommendationrdquo Knowledge and InformationSystems vol 56 no 3 pp 559ndash579 2017

[22] R Srivastava S Hingmire G K Palshikar S Chaurasia andD A Csrs ldquoA context and sequence aware recommendationsystemrdquo in Proceedings of 8th Annual Meeting of the Forum onInformation Retrieval Evaluation pp 8ndash15 Kolkata IndiaDecember 2016

[23] V Subramaniyaswamy V Vijayakumar R Logesh andV Indragandhi ldquoIntelligent travel recommendation systemby mining attributes from community contributed photosrdquoProcedia Computer Science vol 50 pp 447ndash455 2015

[24] L Cao J Luo A C Gallagher X Jin J Han and T S HuangldquoA worldwide tourism recommendation system based ongeotaggedweb photosrdquo in Proceedings of ICASSP pp 2274ndash2277 Dallas Texas USA March 2010

[25] L Guo J Shao K L Tan and Y Y Wheretogo ldquoPersonalizedtravel recommendation for individuals and groupsrdquo in Pro-ceedings of IEEE 15th International Conference onMobile DataManagement vol 1 pp 49ndash58 Brisbane Australia July 2014

[26] I Memon L Chen A Majid M Lv I Hussain and G ChenldquoTravel recommendation using geo-tagged photos in socialmedia for touristrdquoWireless Personal Communications vol 80no 4 pp 1347ndash1362 2015

Mobile Information Systems 11

[27] Z Xu L Chen and G Chen ldquoTopic based context-awaretravel recommendation method exploiting geotaggedphotosrdquo Neurocomputing vol 155 pp 99ndash107 2015

[28] B Dhake S S Lomte R A Auti Y R Nagargoje and B PatilldquoLARS an efficient and scalable location-awarerdquo In-ternational Journal of Scientific Research And Educationvol 2 p 11 2014

[29] M Sarwat J J Levandoski A Eldawy and M F MokbelldquoLARSlowast an efficient and scalable location-aware recom-mender systemrdquo IEEE Transactions on Knowledge and DataEngineering vol 26 no 6 pp 1384ndash1399 2014

[30] P V Krishna S Misra D Joshi andM S Obaidat ldquoLearningautomata based sentiment analysis for recommender systemon cloudrdquo in Proceedings of International Conference onComputer Information and Telecommunication Systems(CITS) pp 1ndash5 IEEE Athens Greece May 2013

[31] M Mordacchini A Passarella M Conti et al ldquoCrowd-sourcing through cognitive opportunistic networksrdquo ACMTransactions on Autonomous and Adaptive Systems vol 10no 2 pp 1ndash29 2015

[32] Y Zheng L Zhang Z Ma X Xie and W Y Ma ldquoRec-ommending friends and locations based on individual loca-tion historyrdquo ACM Transactions on the Web vol 5 no 1pp 1ndash44 2011

[33] VW Zheng Y Zheng X Xie and Q Yang ldquoTowards mobileintelligence learning from GPS history data for collaborativerecommendationrdquo Artificial Intelligence vol 184ndash185pp 17ndash37 2012

[34] M Korakakis E Spyrou P Mylonas and S J PerantonisldquoExploiting social media information toward a context-awarerecommendation systemrdquo Social Network Analysis andMining vol 7 no 1 p 42 2017

[35] Q Li Y Zheng X Xie Y Chen W Liu and W Y MaldquoMining user similarity based on location historyrdquo in Pro-ceedings of 16th ACM SIGSPATIAL International Conferenceon Advances in Geographic Information Systems p 34 ACMNew York NY USA November 2008

[36] M Ye P Yin and W C Lee ldquoLocation recommendation forlocation-based social networksrdquo in Proceedings of 18th SIG-SPATIAL International Conference on Advances in GeographicInformation Systems pp 458ndash461 ACM San Jose CA USANovember 2010

[37] T Hasegawa and T Hayashi ldquoCollaborative filtering basedspot recommendation seamlessly available in home and awayareasrdquo in Proceedings of the IEEEACIS 12th InternationalConference on Computer and Information Science (ICIS)pp 547-548 IEEE Niigata Japan June 2013

[38] J J Levandoski M Sarwat A Eldawy and M F MokbelldquoLars a location-aware recommender systemrdquo in Proceedingsof IEEE 28th International Conference on Data Engineering(ICDE) pp 450ndash461 IEEE Arlington VA USA April 2012

[39] Y Takeuchi and M Sugimoto ldquoCityvoyager an outdoorrecommendation system based on user location historyrdquo inUbiquitous Intelligence and Computing pp 625ndash636Springer-Verlag Berlin Germany 2006

[40] S Knoch A Chapko A Emrich D Werth and P Loos ldquoAcontext-aware running route recommender learning fromuser histories using artificial neural networksrdquo in Proceedingsof 23rd International Workshop on Database and ExpertSystems Applications (DEXA) pp 106ndash110 IEEE ViennaAustria September 2012

[41] M Saraee S Khan and S Yamaner ldquoData mining approachto implement a recommendation system for electronic tourguidesrdquo in Proceedings of 2005 International Conference on

E-Business Enterprise Information Systems E-Governmentand Outsourcing EEE 2005 pp 215ndash218 CSREA PressLas Vegas NV USA June 2005

[42] R Katarya and O P Verma ldquoRestaurant recommendersystem based on psychographic and demographic factors inmobile environmentrdquo in Proceedings of International Con-ference on Green Computing and Internet of Dings (ICG-CIoT) pp 907ndash912 Uttar Pradesh India October 2015

[43] R Katarya and O P Verma ldquoRecent developments in af-fective recommender systemsrdquo Physica A Statistical Me-chanics and its Applications vol 461 pp 182ndash190 2016

[44] R Katarya ldquoA systematic review of group recommendersystems techniquesrdquo in Proceedings of International Confer-ence on Intelligent Sustainable Systems (ICISS) pp 425ndash428Tirupur India December 2017

[45] R Katarya O P Verma and I Jain ldquoUser behaviour analysisin context-aware recommender system using hybrid filteringapproachrdquo in Proceedings of 4th International Conference onComputer and Communication Technology (ICCCT)pp 222ndash227 Allahabad India September 2013

[46] R Katarya M Ranjan and O P Verma ldquoLocation basedrecommender system using enhanced random walk modelrdquoin Proceedings of Fourth International Conference on ParallelDistributed and Grid Computing (PDGC) pp 33ndash37Waknaghat India December 2016

[47] K Waga A Tabarcea and P Franti ldquoContext aware rec-ommendation of location-based datardquo in Proceedings of15th International Conference on System Deory Controland Computing (ICSTCC) pp 1ndash6 IEEE Sinaia RomaniaOctober 2011

[48] A Majid L Chen G Chen H T Mirza I Hussain andJ Woodward ldquoA context-aware personalized travel recom-mendation system based on geotagged social media dataminingrdquo International Journal of Geographical InformationScience vol 27 no 4 pp 662ndash684 2013

[49] Q Yuan G Cong Z Ma A Sun and N M)almann ldquoTime-aware point-of-interest recommendationrdquo in Proceedings of36th International ACM SIGIR Conference on Research andDevelopment in Information Retrieval pp 363ndash372 ACMDublin Ireland July 2013

[50] R Baral and T Li ldquoMaps a multi aspect personalized poirecommender systemrdquo in Proceedings of 10th ACM Confer-ence on Recommender Systems pp 281ndash284 ACM BostonMA USA September 2016

[51] P Hiesel M Braunhofer and W Worndl ldquoLearning thepopularity of items for mobile tourist guidesrdquo in Proceedingsof RecTour Cambridge MA USA September 2016

[52] L Yao Q Z Sheng X Wang W E Zhang and Y QinldquoCollaborative location recommendation by integratingmulti-dimensional contextual informationrdquo ACM Trans-actions on Internet Technology vol 18 no 3 pp 1ndash24 2018

[53] C Trattner A Oberegger L Eberhard D Parra andL Marinho ldquoUnderstanding the impact of weather for POIrecommendationsrdquo in Proceedings of RecTour CambridgeMA USA September 2016

[54] Y Zheng R Burke and B Mobasher ldquoDifferential contextrelaxation for context-aware travel recommendationrdquo inProceedings of International Conference on Electronic Com-merce and Web Technologies pp 88ndash99 Springer ViennaAustria September 2012

[55] M J Barranco J M Noguera J Castro and L Martinez ldquoAcontext-aware mobile recommender system based on locationand trajectoryrdquo in Management Intelligent Systems pp 153ndash162 Springer Berlin Heidelberg Germany 2012

12 Mobile Information Systems

[56] N S Savage M Baranski N E Chavez and T Hollerer ldquoIrsquomfeeling loco a location based context aware recommendationsystemrdquo in Advances in Location-Based Services pp 37ndash54Springer Berlin Heidelberg Germany 2012

[57] J L Herlocker J A Konstan L G Terveen and J T RiedlldquoEvaluating collaborative filtering recommender systemsrdquoACM Transactions on Information Systems vol 22 no 1pp 5ndash53 2004

[58] P Lops M De Gemmis and G Semeraro ldquoContent-basedrecommender systems state of the art and trendsrdquo in Rec-ommender Systems Handbook pp 73ndash105 Springer BerlinHeidelberg Germany 2011

[59] Z Lu H Wang N Mamoulis W Tu and D W CheungldquoPersonalized location recommendation by aggregatingmultiple recommenders in diversityrdquo GeoInformatica vol 21no 3 pp 459ndash484 2017

[60] R Burke ldquoHybrid recommender systems survey and ex-perimentsrdquo User Modeling and User-Adapted Interactionvol 12 no 4 pp 331ndash370 2002

[61] L Baltrunas and X Amatriain ldquoTowards time-dependentrecommendation based on implicit feedbackrdquo in Pro-ceedings of Workshop on Context-Aware Recommender Sys-tems (CARSrsquo09) New York NY USA October 2009

[62] X Wang Y L Zhao L Nie et al ldquoSemantic-based locationrecommendation with multimodal venue semanticsrdquo IEEETransactions on Multimedia vol 17 no 3 pp 409ndash4192015

[63] Q Yuan G Cong and A Sun ldquoGraph-based point-of-interest recommendation with geographical and temporalinfluencesrdquo in Proceedings of 23rd ACM International Con-ference on Information and Knowledge Managementpp 659ndash668 Shanghai China November 2014

[64] Y Mao P Yin W Lee and D Lee ldquoExploiting geographicalinfluence for collaborative point-of-interest recommenda-tionrdquo in Proceedings of 34th ACM SI- GIR InternationalConference on Research and Development in InformationRetrieval pp 325ndash334 Beijing China July 2011

[65] S K Gorakala and M Usuelli Building a RecommendationSystem with R Packt Publishing Ltd Birmingham UK 2015

[66] Y Zheng L Zhang X Xie andW Y Ma ldquoMining interestinglocations and travel sequences from GPS trajectoriesrdquo inProceedings of 18th International Conference on World WideWeb pp 791ndash800 Madrid Spain April 2009

[67] G Shani and A Gunawardana ldquoEvaluating recommendationsystemsrdquo in Recommender Systems Handbook pp 257ndash297Springer Berlin Heidelberg Germany 2011

[68] J A Konstan B N Miller D Maltz J L HerlockerL R Gordon and J Riedl ldquoGroupLens applying collabo-rative filtering to Usenet newsrdquo Communications of the ACMvol 40 no 3 pp 77ndash87 1997

[69] M Deshpande and G Karypis ldquoItem-based top-n recom-mendation algorithmsrdquo ACM Transactions on InformationSystems vol 22 no 1 pp 143ndash177 2004

[70] Y Koren R Bell and C Volinsky ldquoMatrix factorizationtechniques for recommender systemsrdquo Computer vol 42no 8 pp 30ndash37 2009

[71] Y Koren ldquoFactorization meets the neighborhood a multi-faceted collaborative filtering modelrdquo in Proceedings of 14thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 426ndash434 Las Vegas NV USAAugust 2008

Mobile Information Systems 13

Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems

Hindawiwwwhindawicom

Volume 2018

International Journal of

ReconfigurableComputing

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawiwwwhindawicom Volume 2018

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational Intelligence and Neuroscience

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018

Human-ComputerInteraction

Advances in

Hindawiwwwhindawicom Volume 2018

Scientic Programming

Submit your manuscripts atwwwhindawicom

snowy and stormy etc) It was observed that some weathercategories showed the same patterns of user behavior)erefore they were discretized under three main cate-gories ldquosunnyrdquo ldquorainyrdquo (all categories including rainy) andldquosnowyrdquo (all categories including snowy) )e percentage ofwhich contextual circumstances the venue is preferred wascalculated by the proportion of visits in the specific categoryover the total visits

310 Development of Recommendation System Developmentsteps of this recommendation system are explained in detailbelow

User similarity values were calculated from the user-venue matrix which presents the ratings of users to thevenues with cosine similarity (Equation (6)) Ratings(ratingucf ) of the user to the other venues were predicted byusing user similarity values (Equation (7))

user_sim urarr

n urarr

m( 1113857 cos urarr

n urarr

m( 1113857 urarr

n urarr

m

urarr

n

u

rarrm

(6)

ratingucf 1113936 user_sim un um( 1113857 times rating um vn( 1113857

1113936 user_sim un um( 1113857 (7)

User similarity values were calculated from the user-category matrix (Table 7) which presents the number ofvenues that a specific user visited for each category withCosine similarity in a similar manner as in Equation (2)

Popularity values were discretized into three cate-gories namely high medium and low according totheir normalized popularity values which were attainedfrom PCA )en the user-popularity preference matrix(Table 8) was generated which keeps the number ofvenues that a specific user visited in each popularitycategory

User similarity values were calculated from the user-popularity matrix which presents the number of venuesthat a specific user visited in each popularity class with

cosine similarity in the similar manner as in Equation(2)

)e user-price preference matrix (Table 9) was alsoconstructed which keeps the number of venues that a spe-cific user visited in each price category User similarityaccording to price preferences was also calculated in thesimilar manner as in Equation (2)

Equation (3) was also used to calculate predicted ratingsusing user similarity which depend on the category(ratingcategory) popularity (ratingpopularity) and price(ratingprice) preferences of users

Venue similarity values were calculated from the user-venue matrix with cosine distance (Equation (8)) Ratings(ratingicf ) of the user to the other venues were also predictedby using venue similarity values (Equation (9))

venue_sim vrarr

n vrarr

m( 1113857 cos vrarr

n vrarr

m( 1113857 vrarr

n vrarr

m

vrarr

n

v

rarrm

(8)

ratingicf 1113936 venue_sim vn vm( 1113857 times rating un vm( 1113857

1113936 venue_sim vn vm( 1113857

(9)

)e venue-context matrix (Table 10) which keeps trackof the contextual features of the venues was prepared )epercentages showing the venue preferences in differentcontextual circumstances were presented in this matrixWith the help of this matrix venue similarities were alsocalculated using Equation (4) Predicted ratings(ratingcontext) are also calculated as in Equation (5) using thecontextual similarity of venues

)e calculation of ratings (ratingdistance) according to thedistance between a venue to the user depends on the as-sumption that if this distance is short then the user will visitthat venue more frequently [63ndash66] Although the users are

more willing to check in at nearby venues to their centersdistance perception of each user is different Optimal co-efficients (A B and n) for power law distribution [49 64 65]were determined to model the willingness of a user to go andcheck in at a place

ratingdistance A + Blowast distancen (10)

In this study a weighted hybridization technique wasused to compute the score of recommended items using allavailable recommendation algorithms Artificial neuralnetwork (ANN) analysis was applied in order to find theoptimal weights for each technique instead of using equalweights for them Inputs to the ANN were the results of

Mobile Information Systems 5

all available recommendation algorithms and the outputto be predicted was the actual rating Final ratings

were calculated by multiplying the ANN weights andratings

Rating w1 lowast ratingucf + w2 lowast ratingicf + w3 lowast ratingdistance+ w4 lowast ratingpopularity + w5 lowast ratingcategory+ w6 lowast ratingprice + w7 lowast ratingcontext

(11)

In order to decide whether to recommend a venue to theuser or not different threshold values were used for eachuser )e threshold value for each user was determined bytaking the average ratings of that user After that if thecalculated rating (Equation (11)) was greater than hisherthreshold then it was considered that the user will like thatvenue )is is the first version of the algorithm and it isnamed as ldquoHybRecSysrdquo

Existing recommender systems do not consider that thepreferences of the users are affected by different contextualcircumstances For instance a user may prefer a venue ona rainy weekday at noon while another user may prefer thesame venue in another context In order to handle this issueour system calculates the probability of visiting a venue ina specific contextual category For instance the followingequation calculates the probability of visit of user i to thevenue j in the mornings

P time morning visitij111386811138681113868111386811138681113874 1113875

1113936kcontextual similarityjk lowast probablity of morningik

1113936kcontextual similarityjk

(12)

For each user-venue pair there are 36 different con-textual circumstances (day weekday weekend time

morning noon evening season hot cold weather

sunny rainy snowy) For each situation probabilities werecalculated and the resulting table was constructed(Table 11)

Table 11 respectively presents user id venue id averagerating of related user the percentage that the user will visitthat venue in that time category the percentage that the userwill visit that venue in that day category the percentage thatthe user will visit that venue in that season category thepercentage that the user will visit that venue in that weathercategory total point from all contextual variables (sum of allpercentages) predicted rating and the final decision (Like)

)e sum of all categories of each contextual variableshould add up to one For instance since the day variable hastwo categories weekdays and weekend if a userrsquos probabilityof visiting a specific venue on a weekday is 06 then theprobability that user will visit the same venue on a weekendhas to be 04 )erefore the sum of the values of all con-textual variables (Contexttotal) may have a maximum valueof 4 )e final decision of whether a user will like a venue ornot depends on two things the predicted rating havinga greater value than the average rating of the user and thetotal context having a value of at least 2 out of 4 )e finalversion of the algorithm was called ldquocontextually

personalized HybRecSysrdquo and the whole developmentprocess is depicted in Figure 1

4 Evaluation of Recommendation System

)e performance evaluation of the proposed system will beexplained in detail in this section For the evaluation ofrecommender systems there are three types of experimentalsettings [67]

(1) Offline experiments in which a precollected datasetof users is used to validate the results

(2) User studies in which a small sample of users areasked to perform several tasks requiring an in-teraction with the recommendation system

(3) Online evaluation trying to evaluate different algo-rithms over online recommender systemwith a smallpercentage of the traffic

41 Offline Experiments )e Contextually personalizedHybRecSys was compared with five algorithms user-basedK-nearest neighborhood (KNN) [68] item-based KNN [69]biased matrix factorization [70] SVD++ [71] andHybRecSys [1] First four algorithms were available in theLibRec a Java library for recommender systems For each

Table 7 Sample of user-category matrix

User id Turkish_rest American_rest Chinese_rest11949 10 0 010147822 0 6 210437542 0 2 0

Table 8 Sample of user-popularity matrix

User id Low_popularity Medium_popularity High_popularity

11949 7 3 010147822 0 5 310437542 2 0 0

Table 9 Sample of user-price matrix

User id 1 2 3 411949 6 3 1 010147822 0 0 3 510437542 2 0 0 0

6 Mobile Information Systems

algorithm the default settings of LibRec were used Fifthalgorithm HybRecSys is the earlier version of ContextuallyPersonalized HybRecSys Other hybrid algorithms definedin the literature could not be included in this study since thesource codes were not available

Precision recall and F-1 measures were used as eval-uation metrics in this study Precision specifies the per-centage of correctly recommended items over totalrecommended items Recall indicates the percentage ofrecommended items over the total number of liked items bythe user)e F1 scoremeasures the accuracy of the system byusing both precision and recall

In order to split the data into training and test sets K-fold (K 10) cross-validation technique was used )edataset was divided into 10 disjoint sets making sure that

each set contains about 10 of the visits of each user One setwas used as the test set and nine sets were used as the trainingset for each fold

Figure 2 shows precision recall and F1 measure of eachalgorithm and it is obvious that Contextually PersonalizedHybRecSys outperforms all other algorithms

According to precision metric HybRecSys user-basedKNN item-based KNN biased matrix factorization andSVD++ follow contextually personalized HybRecSys re-spectively According to recall metric HybRecSys biasedmatrix factorization item-based KNN SVD++ and user-based KNN follow contextually personalized HybRecSysAccording to F-1 measure HybRecSys item-based KNNbiased matrix factorization user-based KNN and SVD++follow contextually personalized HybRecSys respectively

User-venuematrix

User-categorypreferences

matrix

User-popularitypreferences

matrix

User-pricepreferences

matrix

Venue-contextual

feature matrix

UsersVenues

Check-in historyVenue properties

(category popularity price)Contextual data

(weather time frame days season)

Distance

Contextualprobabilities

Ratingcontext

Contexttotal Rating Average rating of a user Recommend2 amp

Ratingprice Ratingpopularity Ratingcategory RatingdistanceRatingicf Ratingucf

Figure 1 Framework of contextually personalized HybRecSys

Table 11 )e final decision table

Uid Vid Av_rating Time Day Season Weather Contexttotal Rating Likeu1 v1 25 Morning 05 Wdays 0 Hot 1 Sunny 1 25 3 Trueu1 v1 25 Morning 05 Wdays 0 Hot 1 Rainy 0 15 3 Falseu1 v1 25 Morning 05 Wdays 0 Hot 1 Snowy 0 15 3 Falseu1 v1 25 Morning 05 Wdays 0 Cold 0 Sunny 1 15 3 False

Table 10 Sample of venue-context matrix

Venue id Hot_season Cold_season W_day W_end Morning Noon Evening Sunny Rainy SnowyVenue1 075 025 068 032 008 017 075 067 025 008Venue2 05 05 05 05 0 1 0 1 0 0Venue3 083 017 096 004 013 052 035 065 035 0

Mobile Information Systems 7

42 User Study for Contextually Personalized HybRecSysAs was indicated in the literature user studies are veryhelpful to understand whether the recommendations areliked by the users and to collect more detailed data about therecommendation system [45 57 67] Although conductinga user study is difficult time-consuming and costly it issuggested to apply them after the offline experiments inorder to validate the results of the offline experiments [67]

43 Steps of the User Study For this study a user study wasconducted on the users in our dataset For this purpose theTwitter account of each user in our dataset was checked tolearn whether their profiles allow to receive direct messagesOut of 1101 users 195 accounts were open for directmessage )ose users were invited to attend our user studyby direct message and a small incentive (a movie ticket) waspromised if they attend this evaluation

Twenty-four users replied to themessage and accepted toattend the study After that the user evaluation occurred inthe following steps

(1) )e algorithm predicted the ratings of that user to allvenues except the venues they visited

(2) Among the results the algorithm recommendedthree top-rated venues for each 24 users )ese userswere asked via Twitter message whether any of therecommended items attracted their attention

(3) Twenty-one users replied to the recommendationsOnly one of them said that none of the venues wassuitable for them Others were interested in at leastone venue among the recommended ones

(4) A survey which also includes the Foursquare links ofrecommended venues was prepared according to theparticipantsrsquo choices and sent to them

(5) )e users were asked to fill out the survey after theyvisit that recommended venue or after examiningthe Foursquare page of the venues

44 Survey Questions of the User Study )ere are 12 ques-tions in the survey )e first question was asked to un-derstand the appreciation of the participants to the

recommended venues It was asked in 5-point Likert scale (1-Not Like At All 2-Not Like 3-Not Sure 4-Like 5-Like VeryMuch)

)e following four questions were asked to measure theappropriateness of the category price popularity and thelocation of the recommended venue for the participant)eywere also asked using a 5-point Likert scale (1-Not Ap-propriate At All 2-Not Appropriate 3-Not Sure 4-Ap-propriate 5-Very Appropriate)

)e following four questions were asked to understandin which contextual circumstances the user would prefer therecommended venue )ese questions were asked as fixedsum scale questions )e participants were asked to dis-tribute a hundred points to the categories of a contextualvariable according to the tendency of the user to visit thatvenue in these categories

)e last two questions were demographic questions)ey were asked to learn the age gender and the educationlevel of the participants

45 Results of the User Study Participantsrsquo average age is 30and the age range varies from 19 to 38 )ere are eightwomen and 12 men in the dataset 12 of the participantsgraduated from high school six of them have a bachelorrsquosdegree and two of them have a masterrsquos degree Table 12presents the answers of the participants to the first fivequestions of the survey

Ninety percent of the participants liked the recom-mended venues Ten percent of them were indecisive aboutwhether they like the recommended venue Ninety percentof the participants thought that the price class of the rec-ommended venue was suitable for them Forty percent of theparticipants thought that the category of recommendedvenues were very appropriate for them while 20 of themthought that the categories were appropriate and theremaining are indecisive Seventy percent of the participantsthought that the popularity class of recommended venueswas appropriate for them and 10 of them thought that thepopularity classes of the venues were very appropriate Onthe other hand 10 were indecisive while another 10thought that the popularity classes of the venues were notappropriate Moreover the address of the recommended

05

04

03

02

01

0User-based

KNNItem-based

KNNBiased matrixfactorization

SVD++ HybRecSys(neural)

ContextualpersonalizedHybRecSys

PrecisionRecallF1-measure

Figure 2 Graph of precision recall and F-1 measures of the algorithms

8 Mobile Information Systems

venue was given to the participants and they were askedwhether the location was appropriate or not 40 of theparticipants said very appropriate and 30 of them saidappropriate while the other 30 said not appropriate

)e following four questions were asked to understandin which contextual circumstances participants will prefer tovisit the recommended venues Table 13 presents the an-swers of each participant to these questions and the pre-dictions which are calculated from the algorithms for eachuser

)e predicted results were compared with the actualanswers of the participants )e evaluation of the recom-mendation was measured with precision recall and F-1measures Table 14 presents the precision recall and F-1scores for 20 participants Precision value is 0282 recallvalue is 0276 and F1-score is 0279

)e participants made some comments about the venuesvia Twitter messages Some of the participants had been tosome of the restaurants and supported the given recom-mendation with the following expressions

Table 12 Survey answers (Q1ndashQ5)

1 () 2 () 3 () 4 () 5 ()1 Rate your liking 10 80 102 Rate the appropriateness of the price range 10 903 Rate the appropriateness of the category of therestaurant 40 20 40

4 Rate the appropriateness of the popularity class ofthe restaurant 10 10 70 10

5 Rate the appropriateness of the location of therestaurant 30 30 40

Table 13 Answers of questions 6-7-8-9 and prediction of the answers

Answers of questions 6-7-8-9 Prediction of the answersM N E Wdays Wend Hot Cold Sunny Rainy Snowy M N E Wdays Wend Hot Cold Sunny Rainy Snowy

1 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 0002 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 0003 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 0004 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 0005 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 0006 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 0007 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 0008 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 0009 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02510 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 00011 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 00012 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 00013 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 00014 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 00015 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 00016 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 00017 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 00018 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 00019 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02520 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 000

Table 14 Precision recall and F-1 measure for user study

Approaches Precision Recall F-1 scoreUser-based KNN 01220 00823 00983Item-based KNN 01154 01107 01130Biased matrix factorization 00893 01200 01031SVD++ 00702 00976 00816HybRecSys 01667 01460 01493Contextually personalized HybRecSys 018 045 025User study results of contextually personalizedHybRecSys 0282 0276 0279

Mobile Information Systems 9

(i) Participant 1 (Gender Male Age 37 EducationHigh School)

(a) ldquoI always prefer going these two restaurants thatyou have recommendedrdquo

(ii) Participant 2 (Gender Male Age 38 EducationHigh School)

(a) ldquoI have already been to one of the restaurantsrdquo

(iii) Participant 3 (Gender Female Age 19 EducationHigh School)

(a) ldquoI have visited one of the restaurants beforerdquo

(iv) Participant 4 (Gender Female Age 38 EducationBachelorrsquos Degree)

(a) ldquoI have already visited venue1 and venue2rdquo

)ese statements demonstrate the ability of the rec-ommender systemrsquos accurate prediction Even participantsthat had not visited the recommended venues before statedthat they like the recommended venues )is result revealsthat the developed system has the diversity novelty andserendipity dimensions

5 Conclusion

In this study a contextually personalized hybrid recom-mendation model was proposed)is model integrates user-based and item-based collaborative filtering content-basedfiltering together with contextual information in order to getrid of the drawbacks of each approach Different datasources were used to collect the data Visiting history of userswas collected from Twitter Venue characteristics (distancecategory popularity and price classification) were collectedfrom Foursquare and contextual information (weatherseason date and time of visits) related to each visit werecollected from Weather Underground website For content-based filtering the variables distance category popularityand price classification that have not been used before in onealgorithm were used in order to determine content-baseduser similarity Weather conditions season date and timeof each visit were used cumulatively as the properties ofvenues and contextual similarities of venues were utilized)e artificial neural network algorithm was applied to de-termine the weights of each algorithm Ratings coming fromdifferent algorithms (user-based CF item-based CFcontent-based filtering and rating calculated from thecontextual similarities of venues) were used as the predictorsof the actual rating Final ratings were calculated by mul-tiplying the weights retrieved from neural network and theratings from different algorithms In addition in order tomake a more accurate recommendation and make therecommender system contextually personalized our systemcalculates the probability of visiting a venue in a specificcontextual category for each user-venue pair)e decision ofrecommendation of a venue to the specific user is madeaccording to two rules If the calculated rating is greater thanthe average rating of the user and the total contextual score isgreater than two then that venue will be recommended to

that user under the specific contextual circumstances )eHybrid system prunes the disadvantages of each approachthat may occur when they are used separately

Contextually Personalized HybRecSys was compared touser-based and item-based KNN biased matrix factorizationand SVD++ )ese algorithms are evaluated according to themetrics which are used for ranking prediction (precisionrecall and F-1 measure) Training and test datasets werecreated using K-fold cross-validation (K 10) techniqueResults show that contextually personalized HybRecSysoutperforms existing four algorithms according to eachevaluation metric Contextually Personalized HybRecSys ef-fectively overcomes the data sparsity problem using venuecategory popularity and price to model user preferences

In addition recommending very similar venues toprevious visits causes overspecialization )is problem isalso reduced by considering the user preferences with dif-ferent aspects and not just being stuck on only the venuecharacteristics Hence the quality of the recommendation isimproved and the recommender system gained diversitynovelty and serendipity dimensions

)e algorithm partially solves the cold start issue whichcan be caused by both a new user and a new item Even ifa new user rates only one venue the algorithm understandsthe user preferences from the characteristics of the venue(category price and popularity) Moreover the algorithmfigures out the contextual circumstances under which theuser prefers that the specific venue )erefore by looking atthese characteristics the algorithmmay recommend a venueto a new user

)e most important feature that distinguishes the de-veloped algorithm from others is the contextual personal-ization Contextually personalized HybRecSys solvesthis issue and recommends a right venue under the rightconditions

At the beginning the collected data size was large butafter the filtration it became moderately small On theother hand this small dataset generated significant resultsthat may be improved with larger datasets As a futurestudy it is planned to apply contextually personalizedHybRecSys on larger datasets Venue opening and closinghours can be checked for better results In order to solvethe cold start problem entirely the system may recom-mend a venue to a new user by looking at the contextualcircumstances and recommend the most preferred venuebased on these contextual circumstances In additioneven if a new venue is added to the system it can berecommended by looking at the content-related charac-teristics )e online evaluation was conducted with only20 people which is a relatively small sample for obtainingstatistically precise and coherent results )erefore moreusers should be reached for future studies Finally thisalgorithm can be embedded into a mobile application andthis mobile application can be marketed in mobile ap-plication stores )erefore its performance can be mea-sured via online experiments As users download and usethe application they will score the recommendation sothat online feedback can be taken and the algorithm can beimproved

10 Mobile Information Systems

Data Availability

)e Twitter and Foursquare data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

)e authors declare that they have no conflicts of interest

Acknowledgments

)is study was funded by Bogazici University ScientificResearch Fund (grant no 11463)

References

[1] A Bozanta and B Kutlu ldquoHybRecSys content-based con-textual hybrid venue recommender systemrdquo Journal of In-formation Science pp 1ndash15 2018

[2] M Sattari I H Toroslu P Karagoz P Symeonidis andY Manolopoulos ldquoExtended feature combination model forrecommendations in location-based mobile servicesrdquoKnowledge and Information Systems vol 44 no 3 pp 629ndash661 2015

[3] H Wang G Li and J Feng ldquoGroup-based personalizedlocation recommendation on social networksrdquo in WebTechnologies and Applications pp 68ndash80 Springer BerlinGermany 2014

[4] H Yin B Cui Y Sun Z Hu and L Chen ldquoLCARS a spatialitem recommender systemrdquo ACM Transactions on In-formation Systems (TOIS) vol 32 no 3 pp 1ndash37 2014

[5] M H Park J H Hong and S B Cho ldquoLocation-basedrecommendation system using bayesian userrsquos preferencemodel in mobile devicesrdquo in Ubiquitous Intelligence andComputing pp 1130ndash1139 Springer Berlin HeidelbergGermany 2007

[6] K Aihara H Koshiba and H Takeda ldquoBehavioral cost-basedrecommendation model for wanderers in townrdquo in Human-Computer Interaction Towards Mobile and IntelligentInteraction Environments pp 271ndash279 Springer BerlinHeidelberg Germany 2011

[7] Z Yu Y Feng H Xu and X Zhou ldquoRecommending travelpackages based on mobile crowdsourced datardquo IEEE Com-munications Magazine vol 52 no 8 pp 56ndash62 2014

[8] H Yin B Cui L Chen Z Hu and C Zhang ldquoModelinglocation-based user rating profiles for personalized recom-mendationrdquo ACM Transactions on Knowledge Discovery fromData vol 9 no 3 pp 1ndash41 2015

[9] D Zhou S M Rahimi and X Wang ldquoSimilarity-basedprobabilistic category-based location recommendation uti-lizing temporal and geographical influencerdquo InternationalJournal of Data Science and Analytics vol 1 no 2 pp 111ndash121 2016

[10] H Zhu E Chen H Xiong K Yu H Cao and J TianldquoMining mobile user preferences for personalized context-aware recommendationrdquo ACM Transactions on IntelligentSystems and Technology vol 5 no 4 pp 1ndash27 2015

[11] Q Fang C Xu M S Hossain and M G Stcaplrs ldquoA spatial-temporal context-aware personalized location recommen-dation systemrdquo ACM Transactions on Intelligent systems andTechnology vol 7 no 4 pp 1ndash30 2016

[12] Q Yuan G Cong K Zhao Z Ma and A Sun ldquoWho wherewhen and what a nonparametric bayesian approach to

context-aware recommendation and search for twitter usersrdquoACM Transactions on Information Systems vol 33 no 1pp 1ndash31 2015

[13] A Gupta and K Singh ldquoLocation based personalized res-taurant recommendation system for mobile environmentsrdquoin Proceedings of International Conference on Advances inComputing Communications and Informatics (ICACCI)pp 507ndash511 IEEE Mysore India August 2013

[14] J Bao Y Zheng and M F Mokbel ldquoLocation-based andpreference-aware recommendation using sparse geo-socialnetworking datardquo in Proceedings of 20th International Con-ference on Advances in Geographic Information Systemspp 199ndash208 Redondo Beach CA USA November 2012

[15] H Yin Y Sun B Cui Z Hu and L Chen ldquoLcars a location-content-aware recommender systemrdquo in Proceedings of 19thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 221ndash229 Chicago IL USAAugust 2013

[16] M H Kuo L C Chen and C W Liang ldquoBuilding andevaluating a location-based service recommendation systemwith a preference adjustment mechanismrdquo Expert Systemswith Applications vol 36 no 2 pp 3543ndash3554 2009

[17] K Shimada H Uehara and T Endo ldquoA comparative study ofpotential-of-interest days on a sightseeing spot recom-menderrdquo in Proceedings of IIAI 3rd International Conferenceon Advanced Applied Informatics (IIAIAAI) pp 555ndash560Kitakyushu Japan August 2014

[18] L Ahmedi K Rrmoku K Sylejmani and D Shabani ldquoAbimodal social network analysis to recommend points ofinterest to touristsrdquo Social Network Analysis and Miningvol 7 no 1 p 14 2017

[19] H Gao J Tang X Hu and H Liu ldquoContent-aware point ofinterest recommendation on location-based social networksrdquoin Proceedings of the AAAI pp 1721ndash1727 Austin TexasUSA January 2015

[20] Z Zheng Y Chen S Chen L Sun and D Chen ldquoLocation-aware POI recommendation for indoor space by exploitingWiFi logsrdquo Mobile Information Systems vol 2017 Article ID9601404 16 pages 2017

[21] W X Zhao N Zhou A Sun J R Wen J Han andE Y Chang ldquoA time-aware trajectory embedding model fornext-location recommendationrdquo Knowledge and InformationSystems vol 56 no 3 pp 559ndash579 2017

[22] R Srivastava S Hingmire G K Palshikar S Chaurasia andD A Csrs ldquoA context and sequence aware recommendationsystemrdquo in Proceedings of 8th Annual Meeting of the Forum onInformation Retrieval Evaluation pp 8ndash15 Kolkata IndiaDecember 2016

[23] V Subramaniyaswamy V Vijayakumar R Logesh andV Indragandhi ldquoIntelligent travel recommendation systemby mining attributes from community contributed photosrdquoProcedia Computer Science vol 50 pp 447ndash455 2015

[24] L Cao J Luo A C Gallagher X Jin J Han and T S HuangldquoA worldwide tourism recommendation system based ongeotaggedweb photosrdquo in Proceedings of ICASSP pp 2274ndash2277 Dallas Texas USA March 2010

[25] L Guo J Shao K L Tan and Y Y Wheretogo ldquoPersonalizedtravel recommendation for individuals and groupsrdquo in Pro-ceedings of IEEE 15th International Conference onMobile DataManagement vol 1 pp 49ndash58 Brisbane Australia July 2014

[26] I Memon L Chen A Majid M Lv I Hussain and G ChenldquoTravel recommendation using geo-tagged photos in socialmedia for touristrdquoWireless Personal Communications vol 80no 4 pp 1347ndash1362 2015

Mobile Information Systems 11

[27] Z Xu L Chen and G Chen ldquoTopic based context-awaretravel recommendation method exploiting geotaggedphotosrdquo Neurocomputing vol 155 pp 99ndash107 2015

[28] B Dhake S S Lomte R A Auti Y R Nagargoje and B PatilldquoLARS an efficient and scalable location-awarerdquo In-ternational Journal of Scientific Research And Educationvol 2 p 11 2014

[29] M Sarwat J J Levandoski A Eldawy and M F MokbelldquoLARSlowast an efficient and scalable location-aware recom-mender systemrdquo IEEE Transactions on Knowledge and DataEngineering vol 26 no 6 pp 1384ndash1399 2014

[30] P V Krishna S Misra D Joshi andM S Obaidat ldquoLearningautomata based sentiment analysis for recommender systemon cloudrdquo in Proceedings of International Conference onComputer Information and Telecommunication Systems(CITS) pp 1ndash5 IEEE Athens Greece May 2013

[31] M Mordacchini A Passarella M Conti et al ldquoCrowd-sourcing through cognitive opportunistic networksrdquo ACMTransactions on Autonomous and Adaptive Systems vol 10no 2 pp 1ndash29 2015

[32] Y Zheng L Zhang Z Ma X Xie and W Y Ma ldquoRec-ommending friends and locations based on individual loca-tion historyrdquo ACM Transactions on the Web vol 5 no 1pp 1ndash44 2011

[33] VW Zheng Y Zheng X Xie and Q Yang ldquoTowards mobileintelligence learning from GPS history data for collaborativerecommendationrdquo Artificial Intelligence vol 184ndash185pp 17ndash37 2012

[34] M Korakakis E Spyrou P Mylonas and S J PerantonisldquoExploiting social media information toward a context-awarerecommendation systemrdquo Social Network Analysis andMining vol 7 no 1 p 42 2017

[35] Q Li Y Zheng X Xie Y Chen W Liu and W Y MaldquoMining user similarity based on location historyrdquo in Pro-ceedings of 16th ACM SIGSPATIAL International Conferenceon Advances in Geographic Information Systems p 34 ACMNew York NY USA November 2008

[36] M Ye P Yin and W C Lee ldquoLocation recommendation forlocation-based social networksrdquo in Proceedings of 18th SIG-SPATIAL International Conference on Advances in GeographicInformation Systems pp 458ndash461 ACM San Jose CA USANovember 2010

[37] T Hasegawa and T Hayashi ldquoCollaborative filtering basedspot recommendation seamlessly available in home and awayareasrdquo in Proceedings of the IEEEACIS 12th InternationalConference on Computer and Information Science (ICIS)pp 547-548 IEEE Niigata Japan June 2013

[38] J J Levandoski M Sarwat A Eldawy and M F MokbelldquoLars a location-aware recommender systemrdquo in Proceedingsof IEEE 28th International Conference on Data Engineering(ICDE) pp 450ndash461 IEEE Arlington VA USA April 2012

[39] Y Takeuchi and M Sugimoto ldquoCityvoyager an outdoorrecommendation system based on user location historyrdquo inUbiquitous Intelligence and Computing pp 625ndash636Springer-Verlag Berlin Germany 2006

[40] S Knoch A Chapko A Emrich D Werth and P Loos ldquoAcontext-aware running route recommender learning fromuser histories using artificial neural networksrdquo in Proceedingsof 23rd International Workshop on Database and ExpertSystems Applications (DEXA) pp 106ndash110 IEEE ViennaAustria September 2012

[41] M Saraee S Khan and S Yamaner ldquoData mining approachto implement a recommendation system for electronic tourguidesrdquo in Proceedings of 2005 International Conference on

E-Business Enterprise Information Systems E-Governmentand Outsourcing EEE 2005 pp 215ndash218 CSREA PressLas Vegas NV USA June 2005

[42] R Katarya and O P Verma ldquoRestaurant recommendersystem based on psychographic and demographic factors inmobile environmentrdquo in Proceedings of International Con-ference on Green Computing and Internet of Dings (ICG-CIoT) pp 907ndash912 Uttar Pradesh India October 2015

[43] R Katarya and O P Verma ldquoRecent developments in af-fective recommender systemsrdquo Physica A Statistical Me-chanics and its Applications vol 461 pp 182ndash190 2016

[44] R Katarya ldquoA systematic review of group recommendersystems techniquesrdquo in Proceedings of International Confer-ence on Intelligent Sustainable Systems (ICISS) pp 425ndash428Tirupur India December 2017

[45] R Katarya O P Verma and I Jain ldquoUser behaviour analysisin context-aware recommender system using hybrid filteringapproachrdquo in Proceedings of 4th International Conference onComputer and Communication Technology (ICCCT)pp 222ndash227 Allahabad India September 2013

[46] R Katarya M Ranjan and O P Verma ldquoLocation basedrecommender system using enhanced random walk modelrdquoin Proceedings of Fourth International Conference on ParallelDistributed and Grid Computing (PDGC) pp 33ndash37Waknaghat India December 2016

[47] K Waga A Tabarcea and P Franti ldquoContext aware rec-ommendation of location-based datardquo in Proceedings of15th International Conference on System Deory Controland Computing (ICSTCC) pp 1ndash6 IEEE Sinaia RomaniaOctober 2011

[48] A Majid L Chen G Chen H T Mirza I Hussain andJ Woodward ldquoA context-aware personalized travel recom-mendation system based on geotagged social media dataminingrdquo International Journal of Geographical InformationScience vol 27 no 4 pp 662ndash684 2013

[49] Q Yuan G Cong Z Ma A Sun and N M)almann ldquoTime-aware point-of-interest recommendationrdquo in Proceedings of36th International ACM SIGIR Conference on Research andDevelopment in Information Retrieval pp 363ndash372 ACMDublin Ireland July 2013

[50] R Baral and T Li ldquoMaps a multi aspect personalized poirecommender systemrdquo in Proceedings of 10th ACM Confer-ence on Recommender Systems pp 281ndash284 ACM BostonMA USA September 2016

[51] P Hiesel M Braunhofer and W Worndl ldquoLearning thepopularity of items for mobile tourist guidesrdquo in Proceedingsof RecTour Cambridge MA USA September 2016

[52] L Yao Q Z Sheng X Wang W E Zhang and Y QinldquoCollaborative location recommendation by integratingmulti-dimensional contextual informationrdquo ACM Trans-actions on Internet Technology vol 18 no 3 pp 1ndash24 2018

[53] C Trattner A Oberegger L Eberhard D Parra andL Marinho ldquoUnderstanding the impact of weather for POIrecommendationsrdquo in Proceedings of RecTour CambridgeMA USA September 2016

[54] Y Zheng R Burke and B Mobasher ldquoDifferential contextrelaxation for context-aware travel recommendationrdquo inProceedings of International Conference on Electronic Com-merce and Web Technologies pp 88ndash99 Springer ViennaAustria September 2012

[55] M J Barranco J M Noguera J Castro and L Martinez ldquoAcontext-aware mobile recommender system based on locationand trajectoryrdquo in Management Intelligent Systems pp 153ndash162 Springer Berlin Heidelberg Germany 2012

12 Mobile Information Systems

[56] N S Savage M Baranski N E Chavez and T Hollerer ldquoIrsquomfeeling loco a location based context aware recommendationsystemrdquo in Advances in Location-Based Services pp 37ndash54Springer Berlin Heidelberg Germany 2012

[57] J L Herlocker J A Konstan L G Terveen and J T RiedlldquoEvaluating collaborative filtering recommender systemsrdquoACM Transactions on Information Systems vol 22 no 1pp 5ndash53 2004

[58] P Lops M De Gemmis and G Semeraro ldquoContent-basedrecommender systems state of the art and trendsrdquo in Rec-ommender Systems Handbook pp 73ndash105 Springer BerlinHeidelberg Germany 2011

[59] Z Lu H Wang N Mamoulis W Tu and D W CheungldquoPersonalized location recommendation by aggregatingmultiple recommenders in diversityrdquo GeoInformatica vol 21no 3 pp 459ndash484 2017

[60] R Burke ldquoHybrid recommender systems survey and ex-perimentsrdquo User Modeling and User-Adapted Interactionvol 12 no 4 pp 331ndash370 2002

[61] L Baltrunas and X Amatriain ldquoTowards time-dependentrecommendation based on implicit feedbackrdquo in Pro-ceedings of Workshop on Context-Aware Recommender Sys-tems (CARSrsquo09) New York NY USA October 2009

[62] X Wang Y L Zhao L Nie et al ldquoSemantic-based locationrecommendation with multimodal venue semanticsrdquo IEEETransactions on Multimedia vol 17 no 3 pp 409ndash4192015

[63] Q Yuan G Cong and A Sun ldquoGraph-based point-of-interest recommendation with geographical and temporalinfluencesrdquo in Proceedings of 23rd ACM International Con-ference on Information and Knowledge Managementpp 659ndash668 Shanghai China November 2014

[64] Y Mao P Yin W Lee and D Lee ldquoExploiting geographicalinfluence for collaborative point-of-interest recommenda-tionrdquo in Proceedings of 34th ACM SI- GIR InternationalConference on Research and Development in InformationRetrieval pp 325ndash334 Beijing China July 2011

[65] S K Gorakala and M Usuelli Building a RecommendationSystem with R Packt Publishing Ltd Birmingham UK 2015

[66] Y Zheng L Zhang X Xie andW Y Ma ldquoMining interestinglocations and travel sequences from GPS trajectoriesrdquo inProceedings of 18th International Conference on World WideWeb pp 791ndash800 Madrid Spain April 2009

[67] G Shani and A Gunawardana ldquoEvaluating recommendationsystemsrdquo in Recommender Systems Handbook pp 257ndash297Springer Berlin Heidelberg Germany 2011

[68] J A Konstan B N Miller D Maltz J L HerlockerL R Gordon and J Riedl ldquoGroupLens applying collabo-rative filtering to Usenet newsrdquo Communications of the ACMvol 40 no 3 pp 77ndash87 1997

[69] M Deshpande and G Karypis ldquoItem-based top-n recom-mendation algorithmsrdquo ACM Transactions on InformationSystems vol 22 no 1 pp 143ndash177 2004

[70] Y Koren R Bell and C Volinsky ldquoMatrix factorizationtechniques for recommender systemsrdquo Computer vol 42no 8 pp 30ndash37 2009

[71] Y Koren ldquoFactorization meets the neighborhood a multi-faceted collaborative filtering modelrdquo in Proceedings of 14thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 426ndash434 Las Vegas NV USAAugust 2008

Mobile Information Systems 13

Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems

Hindawiwwwhindawicom

Volume 2018

International Journal of

ReconfigurableComputing

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawiwwwhindawicom Volume 2018

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational Intelligence and Neuroscience

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018

Human-ComputerInteraction

Advances in

Hindawiwwwhindawicom Volume 2018

Scientic Programming

Submit your manuscripts atwwwhindawicom

all available recommendation algorithms and the outputto be predicted was the actual rating Final ratings

were calculated by multiplying the ANN weights andratings

Rating w1 lowast ratingucf + w2 lowast ratingicf + w3 lowast ratingdistance+ w4 lowast ratingpopularity + w5 lowast ratingcategory+ w6 lowast ratingprice + w7 lowast ratingcontext

(11)

In order to decide whether to recommend a venue to theuser or not different threshold values were used for eachuser )e threshold value for each user was determined bytaking the average ratings of that user After that if thecalculated rating (Equation (11)) was greater than hisherthreshold then it was considered that the user will like thatvenue )is is the first version of the algorithm and it isnamed as ldquoHybRecSysrdquo

Existing recommender systems do not consider that thepreferences of the users are affected by different contextualcircumstances For instance a user may prefer a venue ona rainy weekday at noon while another user may prefer thesame venue in another context In order to handle this issueour system calculates the probability of visiting a venue ina specific contextual category For instance the followingequation calculates the probability of visit of user i to thevenue j in the mornings

P time morning visitij111386811138681113868111386811138681113874 1113875

1113936kcontextual similarityjk lowast probablity of morningik

1113936kcontextual similarityjk

(12)

For each user-venue pair there are 36 different con-textual circumstances (day weekday weekend time

morning noon evening season hot cold weather

sunny rainy snowy) For each situation probabilities werecalculated and the resulting table was constructed(Table 11)

Table 11 respectively presents user id venue id averagerating of related user the percentage that the user will visitthat venue in that time category the percentage that the userwill visit that venue in that day category the percentage thatthe user will visit that venue in that season category thepercentage that the user will visit that venue in that weathercategory total point from all contextual variables (sum of allpercentages) predicted rating and the final decision (Like)

)e sum of all categories of each contextual variableshould add up to one For instance since the day variable hastwo categories weekdays and weekend if a userrsquos probabilityof visiting a specific venue on a weekday is 06 then theprobability that user will visit the same venue on a weekendhas to be 04 )erefore the sum of the values of all con-textual variables (Contexttotal) may have a maximum valueof 4 )e final decision of whether a user will like a venue ornot depends on two things the predicted rating havinga greater value than the average rating of the user and thetotal context having a value of at least 2 out of 4 )e finalversion of the algorithm was called ldquocontextually

personalized HybRecSysrdquo and the whole developmentprocess is depicted in Figure 1

4 Evaluation of Recommendation System

)e performance evaluation of the proposed system will beexplained in detail in this section For the evaluation ofrecommender systems there are three types of experimentalsettings [67]

(1) Offline experiments in which a precollected datasetof users is used to validate the results

(2) User studies in which a small sample of users areasked to perform several tasks requiring an in-teraction with the recommendation system

(3) Online evaluation trying to evaluate different algo-rithms over online recommender systemwith a smallpercentage of the traffic

41 Offline Experiments )e Contextually personalizedHybRecSys was compared with five algorithms user-basedK-nearest neighborhood (KNN) [68] item-based KNN [69]biased matrix factorization [70] SVD++ [71] andHybRecSys [1] First four algorithms were available in theLibRec a Java library for recommender systems For each

Table 7 Sample of user-category matrix

User id Turkish_rest American_rest Chinese_rest11949 10 0 010147822 0 6 210437542 0 2 0

Table 8 Sample of user-popularity matrix

User id Low_popularity Medium_popularity High_popularity

11949 7 3 010147822 0 5 310437542 2 0 0

Table 9 Sample of user-price matrix

User id 1 2 3 411949 6 3 1 010147822 0 0 3 510437542 2 0 0 0

6 Mobile Information Systems

algorithm the default settings of LibRec were used Fifthalgorithm HybRecSys is the earlier version of ContextuallyPersonalized HybRecSys Other hybrid algorithms definedin the literature could not be included in this study since thesource codes were not available

Precision recall and F-1 measures were used as eval-uation metrics in this study Precision specifies the per-centage of correctly recommended items over totalrecommended items Recall indicates the percentage ofrecommended items over the total number of liked items bythe user)e F1 scoremeasures the accuracy of the system byusing both precision and recall

In order to split the data into training and test sets K-fold (K 10) cross-validation technique was used )edataset was divided into 10 disjoint sets making sure that

each set contains about 10 of the visits of each user One setwas used as the test set and nine sets were used as the trainingset for each fold

Figure 2 shows precision recall and F1 measure of eachalgorithm and it is obvious that Contextually PersonalizedHybRecSys outperforms all other algorithms

According to precision metric HybRecSys user-basedKNN item-based KNN biased matrix factorization andSVD++ follow contextually personalized HybRecSys re-spectively According to recall metric HybRecSys biasedmatrix factorization item-based KNN SVD++ and user-based KNN follow contextually personalized HybRecSysAccording to F-1 measure HybRecSys item-based KNNbiased matrix factorization user-based KNN and SVD++follow contextually personalized HybRecSys respectively

User-venuematrix

User-categorypreferences

matrix

User-popularitypreferences

matrix

User-pricepreferences

matrix

Venue-contextual

feature matrix

UsersVenues

Check-in historyVenue properties

(category popularity price)Contextual data

(weather time frame days season)

Distance

Contextualprobabilities

Ratingcontext

Contexttotal Rating Average rating of a user Recommend2 amp

Ratingprice Ratingpopularity Ratingcategory RatingdistanceRatingicf Ratingucf

Figure 1 Framework of contextually personalized HybRecSys

Table 11 )e final decision table

Uid Vid Av_rating Time Day Season Weather Contexttotal Rating Likeu1 v1 25 Morning 05 Wdays 0 Hot 1 Sunny 1 25 3 Trueu1 v1 25 Morning 05 Wdays 0 Hot 1 Rainy 0 15 3 Falseu1 v1 25 Morning 05 Wdays 0 Hot 1 Snowy 0 15 3 Falseu1 v1 25 Morning 05 Wdays 0 Cold 0 Sunny 1 15 3 False

Table 10 Sample of venue-context matrix

Venue id Hot_season Cold_season W_day W_end Morning Noon Evening Sunny Rainy SnowyVenue1 075 025 068 032 008 017 075 067 025 008Venue2 05 05 05 05 0 1 0 1 0 0Venue3 083 017 096 004 013 052 035 065 035 0

Mobile Information Systems 7

42 User Study for Contextually Personalized HybRecSysAs was indicated in the literature user studies are veryhelpful to understand whether the recommendations areliked by the users and to collect more detailed data about therecommendation system [45 57 67] Although conductinga user study is difficult time-consuming and costly it issuggested to apply them after the offline experiments inorder to validate the results of the offline experiments [67]

43 Steps of the User Study For this study a user study wasconducted on the users in our dataset For this purpose theTwitter account of each user in our dataset was checked tolearn whether their profiles allow to receive direct messagesOut of 1101 users 195 accounts were open for directmessage )ose users were invited to attend our user studyby direct message and a small incentive (a movie ticket) waspromised if they attend this evaluation

Twenty-four users replied to themessage and accepted toattend the study After that the user evaluation occurred inthe following steps

(1) )e algorithm predicted the ratings of that user to allvenues except the venues they visited

(2) Among the results the algorithm recommendedthree top-rated venues for each 24 users )ese userswere asked via Twitter message whether any of therecommended items attracted their attention

(3) Twenty-one users replied to the recommendationsOnly one of them said that none of the venues wassuitable for them Others were interested in at leastone venue among the recommended ones

(4) A survey which also includes the Foursquare links ofrecommended venues was prepared according to theparticipantsrsquo choices and sent to them

(5) )e users were asked to fill out the survey after theyvisit that recommended venue or after examiningthe Foursquare page of the venues

44 Survey Questions of the User Study )ere are 12 ques-tions in the survey )e first question was asked to un-derstand the appreciation of the participants to the

recommended venues It was asked in 5-point Likert scale (1-Not Like At All 2-Not Like 3-Not Sure 4-Like 5-Like VeryMuch)

)e following four questions were asked to measure theappropriateness of the category price popularity and thelocation of the recommended venue for the participant)eywere also asked using a 5-point Likert scale (1-Not Ap-propriate At All 2-Not Appropriate 3-Not Sure 4-Ap-propriate 5-Very Appropriate)

)e following four questions were asked to understandin which contextual circumstances the user would prefer therecommended venue )ese questions were asked as fixedsum scale questions )e participants were asked to dis-tribute a hundred points to the categories of a contextualvariable according to the tendency of the user to visit thatvenue in these categories

)e last two questions were demographic questions)ey were asked to learn the age gender and the educationlevel of the participants

45 Results of the User Study Participantsrsquo average age is 30and the age range varies from 19 to 38 )ere are eightwomen and 12 men in the dataset 12 of the participantsgraduated from high school six of them have a bachelorrsquosdegree and two of them have a masterrsquos degree Table 12presents the answers of the participants to the first fivequestions of the survey

Ninety percent of the participants liked the recom-mended venues Ten percent of them were indecisive aboutwhether they like the recommended venue Ninety percentof the participants thought that the price class of the rec-ommended venue was suitable for them Forty percent of theparticipants thought that the category of recommendedvenues were very appropriate for them while 20 of themthought that the categories were appropriate and theremaining are indecisive Seventy percent of the participantsthought that the popularity class of recommended venueswas appropriate for them and 10 of them thought that thepopularity classes of the venues were very appropriate Onthe other hand 10 were indecisive while another 10thought that the popularity classes of the venues were notappropriate Moreover the address of the recommended

05

04

03

02

01

0User-based

KNNItem-based

KNNBiased matrixfactorization

SVD++ HybRecSys(neural)

ContextualpersonalizedHybRecSys

PrecisionRecallF1-measure

Figure 2 Graph of precision recall and F-1 measures of the algorithms

8 Mobile Information Systems

venue was given to the participants and they were askedwhether the location was appropriate or not 40 of theparticipants said very appropriate and 30 of them saidappropriate while the other 30 said not appropriate

)e following four questions were asked to understandin which contextual circumstances participants will prefer tovisit the recommended venues Table 13 presents the an-swers of each participant to these questions and the pre-dictions which are calculated from the algorithms for eachuser

)e predicted results were compared with the actualanswers of the participants )e evaluation of the recom-mendation was measured with precision recall and F-1measures Table 14 presents the precision recall and F-1scores for 20 participants Precision value is 0282 recallvalue is 0276 and F1-score is 0279

)e participants made some comments about the venuesvia Twitter messages Some of the participants had been tosome of the restaurants and supported the given recom-mendation with the following expressions

Table 12 Survey answers (Q1ndashQ5)

1 () 2 () 3 () 4 () 5 ()1 Rate your liking 10 80 102 Rate the appropriateness of the price range 10 903 Rate the appropriateness of the category of therestaurant 40 20 40

4 Rate the appropriateness of the popularity class ofthe restaurant 10 10 70 10

5 Rate the appropriateness of the location of therestaurant 30 30 40

Table 13 Answers of questions 6-7-8-9 and prediction of the answers

Answers of questions 6-7-8-9 Prediction of the answersM N E Wdays Wend Hot Cold Sunny Rainy Snowy M N E Wdays Wend Hot Cold Sunny Rainy Snowy

1 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 0002 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 0003 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 0004 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 0005 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 0006 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 0007 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 0008 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 0009 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02510 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 00011 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 00012 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 00013 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 00014 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 00015 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 00016 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 00017 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 00018 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 00019 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02520 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 000

Table 14 Precision recall and F-1 measure for user study

Approaches Precision Recall F-1 scoreUser-based KNN 01220 00823 00983Item-based KNN 01154 01107 01130Biased matrix factorization 00893 01200 01031SVD++ 00702 00976 00816HybRecSys 01667 01460 01493Contextually personalized HybRecSys 018 045 025User study results of contextually personalizedHybRecSys 0282 0276 0279

Mobile Information Systems 9

(i) Participant 1 (Gender Male Age 37 EducationHigh School)

(a) ldquoI always prefer going these two restaurants thatyou have recommendedrdquo

(ii) Participant 2 (Gender Male Age 38 EducationHigh School)

(a) ldquoI have already been to one of the restaurantsrdquo

(iii) Participant 3 (Gender Female Age 19 EducationHigh School)

(a) ldquoI have visited one of the restaurants beforerdquo

(iv) Participant 4 (Gender Female Age 38 EducationBachelorrsquos Degree)

(a) ldquoI have already visited venue1 and venue2rdquo

)ese statements demonstrate the ability of the rec-ommender systemrsquos accurate prediction Even participantsthat had not visited the recommended venues before statedthat they like the recommended venues )is result revealsthat the developed system has the diversity novelty andserendipity dimensions

5 Conclusion

In this study a contextually personalized hybrid recom-mendation model was proposed)is model integrates user-based and item-based collaborative filtering content-basedfiltering together with contextual information in order to getrid of the drawbacks of each approach Different datasources were used to collect the data Visiting history of userswas collected from Twitter Venue characteristics (distancecategory popularity and price classification) were collectedfrom Foursquare and contextual information (weatherseason date and time of visits) related to each visit werecollected from Weather Underground website For content-based filtering the variables distance category popularityand price classification that have not been used before in onealgorithm were used in order to determine content-baseduser similarity Weather conditions season date and timeof each visit were used cumulatively as the properties ofvenues and contextual similarities of venues were utilized)e artificial neural network algorithm was applied to de-termine the weights of each algorithm Ratings coming fromdifferent algorithms (user-based CF item-based CFcontent-based filtering and rating calculated from thecontextual similarities of venues) were used as the predictorsof the actual rating Final ratings were calculated by mul-tiplying the weights retrieved from neural network and theratings from different algorithms In addition in order tomake a more accurate recommendation and make therecommender system contextually personalized our systemcalculates the probability of visiting a venue in a specificcontextual category for each user-venue pair)e decision ofrecommendation of a venue to the specific user is madeaccording to two rules If the calculated rating is greater thanthe average rating of the user and the total contextual score isgreater than two then that venue will be recommended to

that user under the specific contextual circumstances )eHybrid system prunes the disadvantages of each approachthat may occur when they are used separately

Contextually Personalized HybRecSys was compared touser-based and item-based KNN biased matrix factorizationand SVD++ )ese algorithms are evaluated according to themetrics which are used for ranking prediction (precisionrecall and F-1 measure) Training and test datasets werecreated using K-fold cross-validation (K 10) techniqueResults show that contextually personalized HybRecSysoutperforms existing four algorithms according to eachevaluation metric Contextually Personalized HybRecSys ef-fectively overcomes the data sparsity problem using venuecategory popularity and price to model user preferences

In addition recommending very similar venues toprevious visits causes overspecialization )is problem isalso reduced by considering the user preferences with dif-ferent aspects and not just being stuck on only the venuecharacteristics Hence the quality of the recommendation isimproved and the recommender system gained diversitynovelty and serendipity dimensions

)e algorithm partially solves the cold start issue whichcan be caused by both a new user and a new item Even ifa new user rates only one venue the algorithm understandsthe user preferences from the characteristics of the venue(category price and popularity) Moreover the algorithmfigures out the contextual circumstances under which theuser prefers that the specific venue )erefore by looking atthese characteristics the algorithmmay recommend a venueto a new user

)e most important feature that distinguishes the de-veloped algorithm from others is the contextual personal-ization Contextually personalized HybRecSys solvesthis issue and recommends a right venue under the rightconditions

At the beginning the collected data size was large butafter the filtration it became moderately small On theother hand this small dataset generated significant resultsthat may be improved with larger datasets As a futurestudy it is planned to apply contextually personalizedHybRecSys on larger datasets Venue opening and closinghours can be checked for better results In order to solvethe cold start problem entirely the system may recom-mend a venue to a new user by looking at the contextualcircumstances and recommend the most preferred venuebased on these contextual circumstances In additioneven if a new venue is added to the system it can berecommended by looking at the content-related charac-teristics )e online evaluation was conducted with only20 people which is a relatively small sample for obtainingstatistically precise and coherent results )erefore moreusers should be reached for future studies Finally thisalgorithm can be embedded into a mobile application andthis mobile application can be marketed in mobile ap-plication stores )erefore its performance can be mea-sured via online experiments As users download and usethe application they will score the recommendation sothat online feedback can be taken and the algorithm can beimproved

10 Mobile Information Systems

Data Availability

)e Twitter and Foursquare data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

)e authors declare that they have no conflicts of interest

Acknowledgments

)is study was funded by Bogazici University ScientificResearch Fund (grant no 11463)

References

[1] A Bozanta and B Kutlu ldquoHybRecSys content-based con-textual hybrid venue recommender systemrdquo Journal of In-formation Science pp 1ndash15 2018

[2] M Sattari I H Toroslu P Karagoz P Symeonidis andY Manolopoulos ldquoExtended feature combination model forrecommendations in location-based mobile servicesrdquoKnowledge and Information Systems vol 44 no 3 pp 629ndash661 2015

[3] H Wang G Li and J Feng ldquoGroup-based personalizedlocation recommendation on social networksrdquo in WebTechnologies and Applications pp 68ndash80 Springer BerlinGermany 2014

[4] H Yin B Cui Y Sun Z Hu and L Chen ldquoLCARS a spatialitem recommender systemrdquo ACM Transactions on In-formation Systems (TOIS) vol 32 no 3 pp 1ndash37 2014

[5] M H Park J H Hong and S B Cho ldquoLocation-basedrecommendation system using bayesian userrsquos preferencemodel in mobile devicesrdquo in Ubiquitous Intelligence andComputing pp 1130ndash1139 Springer Berlin HeidelbergGermany 2007

[6] K Aihara H Koshiba and H Takeda ldquoBehavioral cost-basedrecommendation model for wanderers in townrdquo in Human-Computer Interaction Towards Mobile and IntelligentInteraction Environments pp 271ndash279 Springer BerlinHeidelberg Germany 2011

[7] Z Yu Y Feng H Xu and X Zhou ldquoRecommending travelpackages based on mobile crowdsourced datardquo IEEE Com-munications Magazine vol 52 no 8 pp 56ndash62 2014

[8] H Yin B Cui L Chen Z Hu and C Zhang ldquoModelinglocation-based user rating profiles for personalized recom-mendationrdquo ACM Transactions on Knowledge Discovery fromData vol 9 no 3 pp 1ndash41 2015

[9] D Zhou S M Rahimi and X Wang ldquoSimilarity-basedprobabilistic category-based location recommendation uti-lizing temporal and geographical influencerdquo InternationalJournal of Data Science and Analytics vol 1 no 2 pp 111ndash121 2016

[10] H Zhu E Chen H Xiong K Yu H Cao and J TianldquoMining mobile user preferences for personalized context-aware recommendationrdquo ACM Transactions on IntelligentSystems and Technology vol 5 no 4 pp 1ndash27 2015

[11] Q Fang C Xu M S Hossain and M G Stcaplrs ldquoA spatial-temporal context-aware personalized location recommen-dation systemrdquo ACM Transactions on Intelligent systems andTechnology vol 7 no 4 pp 1ndash30 2016

[12] Q Yuan G Cong K Zhao Z Ma and A Sun ldquoWho wherewhen and what a nonparametric bayesian approach to

context-aware recommendation and search for twitter usersrdquoACM Transactions on Information Systems vol 33 no 1pp 1ndash31 2015

[13] A Gupta and K Singh ldquoLocation based personalized res-taurant recommendation system for mobile environmentsrdquoin Proceedings of International Conference on Advances inComputing Communications and Informatics (ICACCI)pp 507ndash511 IEEE Mysore India August 2013

[14] J Bao Y Zheng and M F Mokbel ldquoLocation-based andpreference-aware recommendation using sparse geo-socialnetworking datardquo in Proceedings of 20th International Con-ference on Advances in Geographic Information Systemspp 199ndash208 Redondo Beach CA USA November 2012

[15] H Yin Y Sun B Cui Z Hu and L Chen ldquoLcars a location-content-aware recommender systemrdquo in Proceedings of 19thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 221ndash229 Chicago IL USAAugust 2013

[16] M H Kuo L C Chen and C W Liang ldquoBuilding andevaluating a location-based service recommendation systemwith a preference adjustment mechanismrdquo Expert Systemswith Applications vol 36 no 2 pp 3543ndash3554 2009

[17] K Shimada H Uehara and T Endo ldquoA comparative study ofpotential-of-interest days on a sightseeing spot recom-menderrdquo in Proceedings of IIAI 3rd International Conferenceon Advanced Applied Informatics (IIAIAAI) pp 555ndash560Kitakyushu Japan August 2014

[18] L Ahmedi K Rrmoku K Sylejmani and D Shabani ldquoAbimodal social network analysis to recommend points ofinterest to touristsrdquo Social Network Analysis and Miningvol 7 no 1 p 14 2017

[19] H Gao J Tang X Hu and H Liu ldquoContent-aware point ofinterest recommendation on location-based social networksrdquoin Proceedings of the AAAI pp 1721ndash1727 Austin TexasUSA January 2015

[20] Z Zheng Y Chen S Chen L Sun and D Chen ldquoLocation-aware POI recommendation for indoor space by exploitingWiFi logsrdquo Mobile Information Systems vol 2017 Article ID9601404 16 pages 2017

[21] W X Zhao N Zhou A Sun J R Wen J Han andE Y Chang ldquoA time-aware trajectory embedding model fornext-location recommendationrdquo Knowledge and InformationSystems vol 56 no 3 pp 559ndash579 2017

[22] R Srivastava S Hingmire G K Palshikar S Chaurasia andD A Csrs ldquoA context and sequence aware recommendationsystemrdquo in Proceedings of 8th Annual Meeting of the Forum onInformation Retrieval Evaluation pp 8ndash15 Kolkata IndiaDecember 2016

[23] V Subramaniyaswamy V Vijayakumar R Logesh andV Indragandhi ldquoIntelligent travel recommendation systemby mining attributes from community contributed photosrdquoProcedia Computer Science vol 50 pp 447ndash455 2015

[24] L Cao J Luo A C Gallagher X Jin J Han and T S HuangldquoA worldwide tourism recommendation system based ongeotaggedweb photosrdquo in Proceedings of ICASSP pp 2274ndash2277 Dallas Texas USA March 2010

[25] L Guo J Shao K L Tan and Y Y Wheretogo ldquoPersonalizedtravel recommendation for individuals and groupsrdquo in Pro-ceedings of IEEE 15th International Conference onMobile DataManagement vol 1 pp 49ndash58 Brisbane Australia July 2014

[26] I Memon L Chen A Majid M Lv I Hussain and G ChenldquoTravel recommendation using geo-tagged photos in socialmedia for touristrdquoWireless Personal Communications vol 80no 4 pp 1347ndash1362 2015

Mobile Information Systems 11

[27] Z Xu L Chen and G Chen ldquoTopic based context-awaretravel recommendation method exploiting geotaggedphotosrdquo Neurocomputing vol 155 pp 99ndash107 2015

[28] B Dhake S S Lomte R A Auti Y R Nagargoje and B PatilldquoLARS an efficient and scalable location-awarerdquo In-ternational Journal of Scientific Research And Educationvol 2 p 11 2014

[29] M Sarwat J J Levandoski A Eldawy and M F MokbelldquoLARSlowast an efficient and scalable location-aware recom-mender systemrdquo IEEE Transactions on Knowledge and DataEngineering vol 26 no 6 pp 1384ndash1399 2014

[30] P V Krishna S Misra D Joshi andM S Obaidat ldquoLearningautomata based sentiment analysis for recommender systemon cloudrdquo in Proceedings of International Conference onComputer Information and Telecommunication Systems(CITS) pp 1ndash5 IEEE Athens Greece May 2013

[31] M Mordacchini A Passarella M Conti et al ldquoCrowd-sourcing through cognitive opportunistic networksrdquo ACMTransactions on Autonomous and Adaptive Systems vol 10no 2 pp 1ndash29 2015

[32] Y Zheng L Zhang Z Ma X Xie and W Y Ma ldquoRec-ommending friends and locations based on individual loca-tion historyrdquo ACM Transactions on the Web vol 5 no 1pp 1ndash44 2011

[33] VW Zheng Y Zheng X Xie and Q Yang ldquoTowards mobileintelligence learning from GPS history data for collaborativerecommendationrdquo Artificial Intelligence vol 184ndash185pp 17ndash37 2012

[34] M Korakakis E Spyrou P Mylonas and S J PerantonisldquoExploiting social media information toward a context-awarerecommendation systemrdquo Social Network Analysis andMining vol 7 no 1 p 42 2017

[35] Q Li Y Zheng X Xie Y Chen W Liu and W Y MaldquoMining user similarity based on location historyrdquo in Pro-ceedings of 16th ACM SIGSPATIAL International Conferenceon Advances in Geographic Information Systems p 34 ACMNew York NY USA November 2008

[36] M Ye P Yin and W C Lee ldquoLocation recommendation forlocation-based social networksrdquo in Proceedings of 18th SIG-SPATIAL International Conference on Advances in GeographicInformation Systems pp 458ndash461 ACM San Jose CA USANovember 2010

[37] T Hasegawa and T Hayashi ldquoCollaborative filtering basedspot recommendation seamlessly available in home and awayareasrdquo in Proceedings of the IEEEACIS 12th InternationalConference on Computer and Information Science (ICIS)pp 547-548 IEEE Niigata Japan June 2013

[38] J J Levandoski M Sarwat A Eldawy and M F MokbelldquoLars a location-aware recommender systemrdquo in Proceedingsof IEEE 28th International Conference on Data Engineering(ICDE) pp 450ndash461 IEEE Arlington VA USA April 2012

[39] Y Takeuchi and M Sugimoto ldquoCityvoyager an outdoorrecommendation system based on user location historyrdquo inUbiquitous Intelligence and Computing pp 625ndash636Springer-Verlag Berlin Germany 2006

[40] S Knoch A Chapko A Emrich D Werth and P Loos ldquoAcontext-aware running route recommender learning fromuser histories using artificial neural networksrdquo in Proceedingsof 23rd International Workshop on Database and ExpertSystems Applications (DEXA) pp 106ndash110 IEEE ViennaAustria September 2012

[41] M Saraee S Khan and S Yamaner ldquoData mining approachto implement a recommendation system for electronic tourguidesrdquo in Proceedings of 2005 International Conference on

E-Business Enterprise Information Systems E-Governmentand Outsourcing EEE 2005 pp 215ndash218 CSREA PressLas Vegas NV USA June 2005

[42] R Katarya and O P Verma ldquoRestaurant recommendersystem based on psychographic and demographic factors inmobile environmentrdquo in Proceedings of International Con-ference on Green Computing and Internet of Dings (ICG-CIoT) pp 907ndash912 Uttar Pradesh India October 2015

[43] R Katarya and O P Verma ldquoRecent developments in af-fective recommender systemsrdquo Physica A Statistical Me-chanics and its Applications vol 461 pp 182ndash190 2016

[44] R Katarya ldquoA systematic review of group recommendersystems techniquesrdquo in Proceedings of International Confer-ence on Intelligent Sustainable Systems (ICISS) pp 425ndash428Tirupur India December 2017

[45] R Katarya O P Verma and I Jain ldquoUser behaviour analysisin context-aware recommender system using hybrid filteringapproachrdquo in Proceedings of 4th International Conference onComputer and Communication Technology (ICCCT)pp 222ndash227 Allahabad India September 2013

[46] R Katarya M Ranjan and O P Verma ldquoLocation basedrecommender system using enhanced random walk modelrdquoin Proceedings of Fourth International Conference on ParallelDistributed and Grid Computing (PDGC) pp 33ndash37Waknaghat India December 2016

[47] K Waga A Tabarcea and P Franti ldquoContext aware rec-ommendation of location-based datardquo in Proceedings of15th International Conference on System Deory Controland Computing (ICSTCC) pp 1ndash6 IEEE Sinaia RomaniaOctober 2011

[48] A Majid L Chen G Chen H T Mirza I Hussain andJ Woodward ldquoA context-aware personalized travel recom-mendation system based on geotagged social media dataminingrdquo International Journal of Geographical InformationScience vol 27 no 4 pp 662ndash684 2013

[49] Q Yuan G Cong Z Ma A Sun and N M)almann ldquoTime-aware point-of-interest recommendationrdquo in Proceedings of36th International ACM SIGIR Conference on Research andDevelopment in Information Retrieval pp 363ndash372 ACMDublin Ireland July 2013

[50] R Baral and T Li ldquoMaps a multi aspect personalized poirecommender systemrdquo in Proceedings of 10th ACM Confer-ence on Recommender Systems pp 281ndash284 ACM BostonMA USA September 2016

[51] P Hiesel M Braunhofer and W Worndl ldquoLearning thepopularity of items for mobile tourist guidesrdquo in Proceedingsof RecTour Cambridge MA USA September 2016

[52] L Yao Q Z Sheng X Wang W E Zhang and Y QinldquoCollaborative location recommendation by integratingmulti-dimensional contextual informationrdquo ACM Trans-actions on Internet Technology vol 18 no 3 pp 1ndash24 2018

[53] C Trattner A Oberegger L Eberhard D Parra andL Marinho ldquoUnderstanding the impact of weather for POIrecommendationsrdquo in Proceedings of RecTour CambridgeMA USA September 2016

[54] Y Zheng R Burke and B Mobasher ldquoDifferential contextrelaxation for context-aware travel recommendationrdquo inProceedings of International Conference on Electronic Com-merce and Web Technologies pp 88ndash99 Springer ViennaAustria September 2012

[55] M J Barranco J M Noguera J Castro and L Martinez ldquoAcontext-aware mobile recommender system based on locationand trajectoryrdquo in Management Intelligent Systems pp 153ndash162 Springer Berlin Heidelberg Germany 2012

12 Mobile Information Systems

[56] N S Savage M Baranski N E Chavez and T Hollerer ldquoIrsquomfeeling loco a location based context aware recommendationsystemrdquo in Advances in Location-Based Services pp 37ndash54Springer Berlin Heidelberg Germany 2012

[57] J L Herlocker J A Konstan L G Terveen and J T RiedlldquoEvaluating collaborative filtering recommender systemsrdquoACM Transactions on Information Systems vol 22 no 1pp 5ndash53 2004

[58] P Lops M De Gemmis and G Semeraro ldquoContent-basedrecommender systems state of the art and trendsrdquo in Rec-ommender Systems Handbook pp 73ndash105 Springer BerlinHeidelberg Germany 2011

[59] Z Lu H Wang N Mamoulis W Tu and D W CheungldquoPersonalized location recommendation by aggregatingmultiple recommenders in diversityrdquo GeoInformatica vol 21no 3 pp 459ndash484 2017

[60] R Burke ldquoHybrid recommender systems survey and ex-perimentsrdquo User Modeling and User-Adapted Interactionvol 12 no 4 pp 331ndash370 2002

[61] L Baltrunas and X Amatriain ldquoTowards time-dependentrecommendation based on implicit feedbackrdquo in Pro-ceedings of Workshop on Context-Aware Recommender Sys-tems (CARSrsquo09) New York NY USA October 2009

[62] X Wang Y L Zhao L Nie et al ldquoSemantic-based locationrecommendation with multimodal venue semanticsrdquo IEEETransactions on Multimedia vol 17 no 3 pp 409ndash4192015

[63] Q Yuan G Cong and A Sun ldquoGraph-based point-of-interest recommendation with geographical and temporalinfluencesrdquo in Proceedings of 23rd ACM International Con-ference on Information and Knowledge Managementpp 659ndash668 Shanghai China November 2014

[64] Y Mao P Yin W Lee and D Lee ldquoExploiting geographicalinfluence for collaborative point-of-interest recommenda-tionrdquo in Proceedings of 34th ACM SI- GIR InternationalConference on Research and Development in InformationRetrieval pp 325ndash334 Beijing China July 2011

[65] S K Gorakala and M Usuelli Building a RecommendationSystem with R Packt Publishing Ltd Birmingham UK 2015

[66] Y Zheng L Zhang X Xie andW Y Ma ldquoMining interestinglocations and travel sequences from GPS trajectoriesrdquo inProceedings of 18th International Conference on World WideWeb pp 791ndash800 Madrid Spain April 2009

[67] G Shani and A Gunawardana ldquoEvaluating recommendationsystemsrdquo in Recommender Systems Handbook pp 257ndash297Springer Berlin Heidelberg Germany 2011

[68] J A Konstan B N Miller D Maltz J L HerlockerL R Gordon and J Riedl ldquoGroupLens applying collabo-rative filtering to Usenet newsrdquo Communications of the ACMvol 40 no 3 pp 77ndash87 1997

[69] M Deshpande and G Karypis ldquoItem-based top-n recom-mendation algorithmsrdquo ACM Transactions on InformationSystems vol 22 no 1 pp 143ndash177 2004

[70] Y Koren R Bell and C Volinsky ldquoMatrix factorizationtechniques for recommender systemsrdquo Computer vol 42no 8 pp 30ndash37 2009

[71] Y Koren ldquoFactorization meets the neighborhood a multi-faceted collaborative filtering modelrdquo in Proceedings of 14thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 426ndash434 Las Vegas NV USAAugust 2008

Mobile Information Systems 13

Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems

Hindawiwwwhindawicom

Volume 2018

International Journal of

ReconfigurableComputing

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawiwwwhindawicom Volume 2018

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational Intelligence and Neuroscience

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018

Human-ComputerInteraction

Advances in

Hindawiwwwhindawicom Volume 2018

Scientic Programming

Submit your manuscripts atwwwhindawicom

algorithm the default settings of LibRec were used Fifthalgorithm HybRecSys is the earlier version of ContextuallyPersonalized HybRecSys Other hybrid algorithms definedin the literature could not be included in this study since thesource codes were not available

Precision recall and F-1 measures were used as eval-uation metrics in this study Precision specifies the per-centage of correctly recommended items over totalrecommended items Recall indicates the percentage ofrecommended items over the total number of liked items bythe user)e F1 scoremeasures the accuracy of the system byusing both precision and recall

In order to split the data into training and test sets K-fold (K 10) cross-validation technique was used )edataset was divided into 10 disjoint sets making sure that

each set contains about 10 of the visits of each user One setwas used as the test set and nine sets were used as the trainingset for each fold

Figure 2 shows precision recall and F1 measure of eachalgorithm and it is obvious that Contextually PersonalizedHybRecSys outperforms all other algorithms

According to precision metric HybRecSys user-basedKNN item-based KNN biased matrix factorization andSVD++ follow contextually personalized HybRecSys re-spectively According to recall metric HybRecSys biasedmatrix factorization item-based KNN SVD++ and user-based KNN follow contextually personalized HybRecSysAccording to F-1 measure HybRecSys item-based KNNbiased matrix factorization user-based KNN and SVD++follow contextually personalized HybRecSys respectively

User-venuematrix

User-categorypreferences

matrix

User-popularitypreferences

matrix

User-pricepreferences

matrix

Venue-contextual

feature matrix

UsersVenues

Check-in historyVenue properties

(category popularity price)Contextual data

(weather time frame days season)

Distance

Contextualprobabilities

Ratingcontext

Contexttotal Rating Average rating of a user Recommend2 amp

Ratingprice Ratingpopularity Ratingcategory RatingdistanceRatingicf Ratingucf

Figure 1 Framework of contextually personalized HybRecSys

Table 11 )e final decision table

Uid Vid Av_rating Time Day Season Weather Contexttotal Rating Likeu1 v1 25 Morning 05 Wdays 0 Hot 1 Sunny 1 25 3 Trueu1 v1 25 Morning 05 Wdays 0 Hot 1 Rainy 0 15 3 Falseu1 v1 25 Morning 05 Wdays 0 Hot 1 Snowy 0 15 3 Falseu1 v1 25 Morning 05 Wdays 0 Cold 0 Sunny 1 15 3 False

Table 10 Sample of venue-context matrix

Venue id Hot_season Cold_season W_day W_end Morning Noon Evening Sunny Rainy SnowyVenue1 075 025 068 032 008 017 075 067 025 008Venue2 05 05 05 05 0 1 0 1 0 0Venue3 083 017 096 004 013 052 035 065 035 0

Mobile Information Systems 7

42 User Study for Contextually Personalized HybRecSysAs was indicated in the literature user studies are veryhelpful to understand whether the recommendations areliked by the users and to collect more detailed data about therecommendation system [45 57 67] Although conductinga user study is difficult time-consuming and costly it issuggested to apply them after the offline experiments inorder to validate the results of the offline experiments [67]

43 Steps of the User Study For this study a user study wasconducted on the users in our dataset For this purpose theTwitter account of each user in our dataset was checked tolearn whether their profiles allow to receive direct messagesOut of 1101 users 195 accounts were open for directmessage )ose users were invited to attend our user studyby direct message and a small incentive (a movie ticket) waspromised if they attend this evaluation

Twenty-four users replied to themessage and accepted toattend the study After that the user evaluation occurred inthe following steps

(1) )e algorithm predicted the ratings of that user to allvenues except the venues they visited

(2) Among the results the algorithm recommendedthree top-rated venues for each 24 users )ese userswere asked via Twitter message whether any of therecommended items attracted their attention

(3) Twenty-one users replied to the recommendationsOnly one of them said that none of the venues wassuitable for them Others were interested in at leastone venue among the recommended ones

(4) A survey which also includes the Foursquare links ofrecommended venues was prepared according to theparticipantsrsquo choices and sent to them

(5) )e users were asked to fill out the survey after theyvisit that recommended venue or after examiningthe Foursquare page of the venues

44 Survey Questions of the User Study )ere are 12 ques-tions in the survey )e first question was asked to un-derstand the appreciation of the participants to the

recommended venues It was asked in 5-point Likert scale (1-Not Like At All 2-Not Like 3-Not Sure 4-Like 5-Like VeryMuch)

)e following four questions were asked to measure theappropriateness of the category price popularity and thelocation of the recommended venue for the participant)eywere also asked using a 5-point Likert scale (1-Not Ap-propriate At All 2-Not Appropriate 3-Not Sure 4-Ap-propriate 5-Very Appropriate)

)e following four questions were asked to understandin which contextual circumstances the user would prefer therecommended venue )ese questions were asked as fixedsum scale questions )e participants were asked to dis-tribute a hundred points to the categories of a contextualvariable according to the tendency of the user to visit thatvenue in these categories

)e last two questions were demographic questions)ey were asked to learn the age gender and the educationlevel of the participants

45 Results of the User Study Participantsrsquo average age is 30and the age range varies from 19 to 38 )ere are eightwomen and 12 men in the dataset 12 of the participantsgraduated from high school six of them have a bachelorrsquosdegree and two of them have a masterrsquos degree Table 12presents the answers of the participants to the first fivequestions of the survey

Ninety percent of the participants liked the recom-mended venues Ten percent of them were indecisive aboutwhether they like the recommended venue Ninety percentof the participants thought that the price class of the rec-ommended venue was suitable for them Forty percent of theparticipants thought that the category of recommendedvenues were very appropriate for them while 20 of themthought that the categories were appropriate and theremaining are indecisive Seventy percent of the participantsthought that the popularity class of recommended venueswas appropriate for them and 10 of them thought that thepopularity classes of the venues were very appropriate Onthe other hand 10 were indecisive while another 10thought that the popularity classes of the venues were notappropriate Moreover the address of the recommended

05

04

03

02

01

0User-based

KNNItem-based

KNNBiased matrixfactorization

SVD++ HybRecSys(neural)

ContextualpersonalizedHybRecSys

PrecisionRecallF1-measure

Figure 2 Graph of precision recall and F-1 measures of the algorithms

8 Mobile Information Systems

venue was given to the participants and they were askedwhether the location was appropriate or not 40 of theparticipants said very appropriate and 30 of them saidappropriate while the other 30 said not appropriate

)e following four questions were asked to understandin which contextual circumstances participants will prefer tovisit the recommended venues Table 13 presents the an-swers of each participant to these questions and the pre-dictions which are calculated from the algorithms for eachuser

)e predicted results were compared with the actualanswers of the participants )e evaluation of the recom-mendation was measured with precision recall and F-1measures Table 14 presents the precision recall and F-1scores for 20 participants Precision value is 0282 recallvalue is 0276 and F1-score is 0279

)e participants made some comments about the venuesvia Twitter messages Some of the participants had been tosome of the restaurants and supported the given recom-mendation with the following expressions

Table 12 Survey answers (Q1ndashQ5)

1 () 2 () 3 () 4 () 5 ()1 Rate your liking 10 80 102 Rate the appropriateness of the price range 10 903 Rate the appropriateness of the category of therestaurant 40 20 40

4 Rate the appropriateness of the popularity class ofthe restaurant 10 10 70 10

5 Rate the appropriateness of the location of therestaurant 30 30 40

Table 13 Answers of questions 6-7-8-9 and prediction of the answers

Answers of questions 6-7-8-9 Prediction of the answersM N E Wdays Wend Hot Cold Sunny Rainy Snowy M N E Wdays Wend Hot Cold Sunny Rainy Snowy

1 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 0002 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 0003 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 0004 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 0005 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 0006 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 0007 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 0008 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 0009 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02510 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 00011 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 00012 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 00013 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 00014 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 00015 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 00016 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 00017 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 00018 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 00019 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02520 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 000

Table 14 Precision recall and F-1 measure for user study

Approaches Precision Recall F-1 scoreUser-based KNN 01220 00823 00983Item-based KNN 01154 01107 01130Biased matrix factorization 00893 01200 01031SVD++ 00702 00976 00816HybRecSys 01667 01460 01493Contextually personalized HybRecSys 018 045 025User study results of contextually personalizedHybRecSys 0282 0276 0279

Mobile Information Systems 9

(i) Participant 1 (Gender Male Age 37 EducationHigh School)

(a) ldquoI always prefer going these two restaurants thatyou have recommendedrdquo

(ii) Participant 2 (Gender Male Age 38 EducationHigh School)

(a) ldquoI have already been to one of the restaurantsrdquo

(iii) Participant 3 (Gender Female Age 19 EducationHigh School)

(a) ldquoI have visited one of the restaurants beforerdquo

(iv) Participant 4 (Gender Female Age 38 EducationBachelorrsquos Degree)

(a) ldquoI have already visited venue1 and venue2rdquo

)ese statements demonstrate the ability of the rec-ommender systemrsquos accurate prediction Even participantsthat had not visited the recommended venues before statedthat they like the recommended venues )is result revealsthat the developed system has the diversity novelty andserendipity dimensions

5 Conclusion

In this study a contextually personalized hybrid recom-mendation model was proposed)is model integrates user-based and item-based collaborative filtering content-basedfiltering together with contextual information in order to getrid of the drawbacks of each approach Different datasources were used to collect the data Visiting history of userswas collected from Twitter Venue characteristics (distancecategory popularity and price classification) were collectedfrom Foursquare and contextual information (weatherseason date and time of visits) related to each visit werecollected from Weather Underground website For content-based filtering the variables distance category popularityand price classification that have not been used before in onealgorithm were used in order to determine content-baseduser similarity Weather conditions season date and timeof each visit were used cumulatively as the properties ofvenues and contextual similarities of venues were utilized)e artificial neural network algorithm was applied to de-termine the weights of each algorithm Ratings coming fromdifferent algorithms (user-based CF item-based CFcontent-based filtering and rating calculated from thecontextual similarities of venues) were used as the predictorsof the actual rating Final ratings were calculated by mul-tiplying the weights retrieved from neural network and theratings from different algorithms In addition in order tomake a more accurate recommendation and make therecommender system contextually personalized our systemcalculates the probability of visiting a venue in a specificcontextual category for each user-venue pair)e decision ofrecommendation of a venue to the specific user is madeaccording to two rules If the calculated rating is greater thanthe average rating of the user and the total contextual score isgreater than two then that venue will be recommended to

that user under the specific contextual circumstances )eHybrid system prunes the disadvantages of each approachthat may occur when they are used separately

Contextually Personalized HybRecSys was compared touser-based and item-based KNN biased matrix factorizationand SVD++ )ese algorithms are evaluated according to themetrics which are used for ranking prediction (precisionrecall and F-1 measure) Training and test datasets werecreated using K-fold cross-validation (K 10) techniqueResults show that contextually personalized HybRecSysoutperforms existing four algorithms according to eachevaluation metric Contextually Personalized HybRecSys ef-fectively overcomes the data sparsity problem using venuecategory popularity and price to model user preferences

In addition recommending very similar venues toprevious visits causes overspecialization )is problem isalso reduced by considering the user preferences with dif-ferent aspects and not just being stuck on only the venuecharacteristics Hence the quality of the recommendation isimproved and the recommender system gained diversitynovelty and serendipity dimensions

)e algorithm partially solves the cold start issue whichcan be caused by both a new user and a new item Even ifa new user rates only one venue the algorithm understandsthe user preferences from the characteristics of the venue(category price and popularity) Moreover the algorithmfigures out the contextual circumstances under which theuser prefers that the specific venue )erefore by looking atthese characteristics the algorithmmay recommend a venueto a new user

)e most important feature that distinguishes the de-veloped algorithm from others is the contextual personal-ization Contextually personalized HybRecSys solvesthis issue and recommends a right venue under the rightconditions

At the beginning the collected data size was large butafter the filtration it became moderately small On theother hand this small dataset generated significant resultsthat may be improved with larger datasets As a futurestudy it is planned to apply contextually personalizedHybRecSys on larger datasets Venue opening and closinghours can be checked for better results In order to solvethe cold start problem entirely the system may recom-mend a venue to a new user by looking at the contextualcircumstances and recommend the most preferred venuebased on these contextual circumstances In additioneven if a new venue is added to the system it can berecommended by looking at the content-related charac-teristics )e online evaluation was conducted with only20 people which is a relatively small sample for obtainingstatistically precise and coherent results )erefore moreusers should be reached for future studies Finally thisalgorithm can be embedded into a mobile application andthis mobile application can be marketed in mobile ap-plication stores )erefore its performance can be mea-sured via online experiments As users download and usethe application they will score the recommendation sothat online feedback can be taken and the algorithm can beimproved

10 Mobile Information Systems

Data Availability

)e Twitter and Foursquare data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

)e authors declare that they have no conflicts of interest

Acknowledgments

)is study was funded by Bogazici University ScientificResearch Fund (grant no 11463)

References

[1] A Bozanta and B Kutlu ldquoHybRecSys content-based con-textual hybrid venue recommender systemrdquo Journal of In-formation Science pp 1ndash15 2018

[2] M Sattari I H Toroslu P Karagoz P Symeonidis andY Manolopoulos ldquoExtended feature combination model forrecommendations in location-based mobile servicesrdquoKnowledge and Information Systems vol 44 no 3 pp 629ndash661 2015

[3] H Wang G Li and J Feng ldquoGroup-based personalizedlocation recommendation on social networksrdquo in WebTechnologies and Applications pp 68ndash80 Springer BerlinGermany 2014

[4] H Yin B Cui Y Sun Z Hu and L Chen ldquoLCARS a spatialitem recommender systemrdquo ACM Transactions on In-formation Systems (TOIS) vol 32 no 3 pp 1ndash37 2014

[5] M H Park J H Hong and S B Cho ldquoLocation-basedrecommendation system using bayesian userrsquos preferencemodel in mobile devicesrdquo in Ubiquitous Intelligence andComputing pp 1130ndash1139 Springer Berlin HeidelbergGermany 2007

[6] K Aihara H Koshiba and H Takeda ldquoBehavioral cost-basedrecommendation model for wanderers in townrdquo in Human-Computer Interaction Towards Mobile and IntelligentInteraction Environments pp 271ndash279 Springer BerlinHeidelberg Germany 2011

[7] Z Yu Y Feng H Xu and X Zhou ldquoRecommending travelpackages based on mobile crowdsourced datardquo IEEE Com-munications Magazine vol 52 no 8 pp 56ndash62 2014

[8] H Yin B Cui L Chen Z Hu and C Zhang ldquoModelinglocation-based user rating profiles for personalized recom-mendationrdquo ACM Transactions on Knowledge Discovery fromData vol 9 no 3 pp 1ndash41 2015

[9] D Zhou S M Rahimi and X Wang ldquoSimilarity-basedprobabilistic category-based location recommendation uti-lizing temporal and geographical influencerdquo InternationalJournal of Data Science and Analytics vol 1 no 2 pp 111ndash121 2016

[10] H Zhu E Chen H Xiong K Yu H Cao and J TianldquoMining mobile user preferences for personalized context-aware recommendationrdquo ACM Transactions on IntelligentSystems and Technology vol 5 no 4 pp 1ndash27 2015

[11] Q Fang C Xu M S Hossain and M G Stcaplrs ldquoA spatial-temporal context-aware personalized location recommen-dation systemrdquo ACM Transactions on Intelligent systems andTechnology vol 7 no 4 pp 1ndash30 2016

[12] Q Yuan G Cong K Zhao Z Ma and A Sun ldquoWho wherewhen and what a nonparametric bayesian approach to

context-aware recommendation and search for twitter usersrdquoACM Transactions on Information Systems vol 33 no 1pp 1ndash31 2015

[13] A Gupta and K Singh ldquoLocation based personalized res-taurant recommendation system for mobile environmentsrdquoin Proceedings of International Conference on Advances inComputing Communications and Informatics (ICACCI)pp 507ndash511 IEEE Mysore India August 2013

[14] J Bao Y Zheng and M F Mokbel ldquoLocation-based andpreference-aware recommendation using sparse geo-socialnetworking datardquo in Proceedings of 20th International Con-ference on Advances in Geographic Information Systemspp 199ndash208 Redondo Beach CA USA November 2012

[15] H Yin Y Sun B Cui Z Hu and L Chen ldquoLcars a location-content-aware recommender systemrdquo in Proceedings of 19thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 221ndash229 Chicago IL USAAugust 2013

[16] M H Kuo L C Chen and C W Liang ldquoBuilding andevaluating a location-based service recommendation systemwith a preference adjustment mechanismrdquo Expert Systemswith Applications vol 36 no 2 pp 3543ndash3554 2009

[17] K Shimada H Uehara and T Endo ldquoA comparative study ofpotential-of-interest days on a sightseeing spot recom-menderrdquo in Proceedings of IIAI 3rd International Conferenceon Advanced Applied Informatics (IIAIAAI) pp 555ndash560Kitakyushu Japan August 2014

[18] L Ahmedi K Rrmoku K Sylejmani and D Shabani ldquoAbimodal social network analysis to recommend points ofinterest to touristsrdquo Social Network Analysis and Miningvol 7 no 1 p 14 2017

[19] H Gao J Tang X Hu and H Liu ldquoContent-aware point ofinterest recommendation on location-based social networksrdquoin Proceedings of the AAAI pp 1721ndash1727 Austin TexasUSA January 2015

[20] Z Zheng Y Chen S Chen L Sun and D Chen ldquoLocation-aware POI recommendation for indoor space by exploitingWiFi logsrdquo Mobile Information Systems vol 2017 Article ID9601404 16 pages 2017

[21] W X Zhao N Zhou A Sun J R Wen J Han andE Y Chang ldquoA time-aware trajectory embedding model fornext-location recommendationrdquo Knowledge and InformationSystems vol 56 no 3 pp 559ndash579 2017

[22] R Srivastava S Hingmire G K Palshikar S Chaurasia andD A Csrs ldquoA context and sequence aware recommendationsystemrdquo in Proceedings of 8th Annual Meeting of the Forum onInformation Retrieval Evaluation pp 8ndash15 Kolkata IndiaDecember 2016

[23] V Subramaniyaswamy V Vijayakumar R Logesh andV Indragandhi ldquoIntelligent travel recommendation systemby mining attributes from community contributed photosrdquoProcedia Computer Science vol 50 pp 447ndash455 2015

[24] L Cao J Luo A C Gallagher X Jin J Han and T S HuangldquoA worldwide tourism recommendation system based ongeotaggedweb photosrdquo in Proceedings of ICASSP pp 2274ndash2277 Dallas Texas USA March 2010

[25] L Guo J Shao K L Tan and Y Y Wheretogo ldquoPersonalizedtravel recommendation for individuals and groupsrdquo in Pro-ceedings of IEEE 15th International Conference onMobile DataManagement vol 1 pp 49ndash58 Brisbane Australia July 2014

[26] I Memon L Chen A Majid M Lv I Hussain and G ChenldquoTravel recommendation using geo-tagged photos in socialmedia for touristrdquoWireless Personal Communications vol 80no 4 pp 1347ndash1362 2015

Mobile Information Systems 11

[27] Z Xu L Chen and G Chen ldquoTopic based context-awaretravel recommendation method exploiting geotaggedphotosrdquo Neurocomputing vol 155 pp 99ndash107 2015

[28] B Dhake S S Lomte R A Auti Y R Nagargoje and B PatilldquoLARS an efficient and scalable location-awarerdquo In-ternational Journal of Scientific Research And Educationvol 2 p 11 2014

[29] M Sarwat J J Levandoski A Eldawy and M F MokbelldquoLARSlowast an efficient and scalable location-aware recom-mender systemrdquo IEEE Transactions on Knowledge and DataEngineering vol 26 no 6 pp 1384ndash1399 2014

[30] P V Krishna S Misra D Joshi andM S Obaidat ldquoLearningautomata based sentiment analysis for recommender systemon cloudrdquo in Proceedings of International Conference onComputer Information and Telecommunication Systems(CITS) pp 1ndash5 IEEE Athens Greece May 2013

[31] M Mordacchini A Passarella M Conti et al ldquoCrowd-sourcing through cognitive opportunistic networksrdquo ACMTransactions on Autonomous and Adaptive Systems vol 10no 2 pp 1ndash29 2015

[32] Y Zheng L Zhang Z Ma X Xie and W Y Ma ldquoRec-ommending friends and locations based on individual loca-tion historyrdquo ACM Transactions on the Web vol 5 no 1pp 1ndash44 2011

[33] VW Zheng Y Zheng X Xie and Q Yang ldquoTowards mobileintelligence learning from GPS history data for collaborativerecommendationrdquo Artificial Intelligence vol 184ndash185pp 17ndash37 2012

[34] M Korakakis E Spyrou P Mylonas and S J PerantonisldquoExploiting social media information toward a context-awarerecommendation systemrdquo Social Network Analysis andMining vol 7 no 1 p 42 2017

[35] Q Li Y Zheng X Xie Y Chen W Liu and W Y MaldquoMining user similarity based on location historyrdquo in Pro-ceedings of 16th ACM SIGSPATIAL International Conferenceon Advances in Geographic Information Systems p 34 ACMNew York NY USA November 2008

[36] M Ye P Yin and W C Lee ldquoLocation recommendation forlocation-based social networksrdquo in Proceedings of 18th SIG-SPATIAL International Conference on Advances in GeographicInformation Systems pp 458ndash461 ACM San Jose CA USANovember 2010

[37] T Hasegawa and T Hayashi ldquoCollaborative filtering basedspot recommendation seamlessly available in home and awayareasrdquo in Proceedings of the IEEEACIS 12th InternationalConference on Computer and Information Science (ICIS)pp 547-548 IEEE Niigata Japan June 2013

[38] J J Levandoski M Sarwat A Eldawy and M F MokbelldquoLars a location-aware recommender systemrdquo in Proceedingsof IEEE 28th International Conference on Data Engineering(ICDE) pp 450ndash461 IEEE Arlington VA USA April 2012

[39] Y Takeuchi and M Sugimoto ldquoCityvoyager an outdoorrecommendation system based on user location historyrdquo inUbiquitous Intelligence and Computing pp 625ndash636Springer-Verlag Berlin Germany 2006

[40] S Knoch A Chapko A Emrich D Werth and P Loos ldquoAcontext-aware running route recommender learning fromuser histories using artificial neural networksrdquo in Proceedingsof 23rd International Workshop on Database and ExpertSystems Applications (DEXA) pp 106ndash110 IEEE ViennaAustria September 2012

[41] M Saraee S Khan and S Yamaner ldquoData mining approachto implement a recommendation system for electronic tourguidesrdquo in Proceedings of 2005 International Conference on

E-Business Enterprise Information Systems E-Governmentand Outsourcing EEE 2005 pp 215ndash218 CSREA PressLas Vegas NV USA June 2005

[42] R Katarya and O P Verma ldquoRestaurant recommendersystem based on psychographic and demographic factors inmobile environmentrdquo in Proceedings of International Con-ference on Green Computing and Internet of Dings (ICG-CIoT) pp 907ndash912 Uttar Pradesh India October 2015

[43] R Katarya and O P Verma ldquoRecent developments in af-fective recommender systemsrdquo Physica A Statistical Me-chanics and its Applications vol 461 pp 182ndash190 2016

[44] R Katarya ldquoA systematic review of group recommendersystems techniquesrdquo in Proceedings of International Confer-ence on Intelligent Sustainable Systems (ICISS) pp 425ndash428Tirupur India December 2017

[45] R Katarya O P Verma and I Jain ldquoUser behaviour analysisin context-aware recommender system using hybrid filteringapproachrdquo in Proceedings of 4th International Conference onComputer and Communication Technology (ICCCT)pp 222ndash227 Allahabad India September 2013

[46] R Katarya M Ranjan and O P Verma ldquoLocation basedrecommender system using enhanced random walk modelrdquoin Proceedings of Fourth International Conference on ParallelDistributed and Grid Computing (PDGC) pp 33ndash37Waknaghat India December 2016

[47] K Waga A Tabarcea and P Franti ldquoContext aware rec-ommendation of location-based datardquo in Proceedings of15th International Conference on System Deory Controland Computing (ICSTCC) pp 1ndash6 IEEE Sinaia RomaniaOctober 2011

[48] A Majid L Chen G Chen H T Mirza I Hussain andJ Woodward ldquoA context-aware personalized travel recom-mendation system based on geotagged social media dataminingrdquo International Journal of Geographical InformationScience vol 27 no 4 pp 662ndash684 2013

[49] Q Yuan G Cong Z Ma A Sun and N M)almann ldquoTime-aware point-of-interest recommendationrdquo in Proceedings of36th International ACM SIGIR Conference on Research andDevelopment in Information Retrieval pp 363ndash372 ACMDublin Ireland July 2013

[50] R Baral and T Li ldquoMaps a multi aspect personalized poirecommender systemrdquo in Proceedings of 10th ACM Confer-ence on Recommender Systems pp 281ndash284 ACM BostonMA USA September 2016

[51] P Hiesel M Braunhofer and W Worndl ldquoLearning thepopularity of items for mobile tourist guidesrdquo in Proceedingsof RecTour Cambridge MA USA September 2016

[52] L Yao Q Z Sheng X Wang W E Zhang and Y QinldquoCollaborative location recommendation by integratingmulti-dimensional contextual informationrdquo ACM Trans-actions on Internet Technology vol 18 no 3 pp 1ndash24 2018

[53] C Trattner A Oberegger L Eberhard D Parra andL Marinho ldquoUnderstanding the impact of weather for POIrecommendationsrdquo in Proceedings of RecTour CambridgeMA USA September 2016

[54] Y Zheng R Burke and B Mobasher ldquoDifferential contextrelaxation for context-aware travel recommendationrdquo inProceedings of International Conference on Electronic Com-merce and Web Technologies pp 88ndash99 Springer ViennaAustria September 2012

[55] M J Barranco J M Noguera J Castro and L Martinez ldquoAcontext-aware mobile recommender system based on locationand trajectoryrdquo in Management Intelligent Systems pp 153ndash162 Springer Berlin Heidelberg Germany 2012

12 Mobile Information Systems

[56] N S Savage M Baranski N E Chavez and T Hollerer ldquoIrsquomfeeling loco a location based context aware recommendationsystemrdquo in Advances in Location-Based Services pp 37ndash54Springer Berlin Heidelberg Germany 2012

[57] J L Herlocker J A Konstan L G Terveen and J T RiedlldquoEvaluating collaborative filtering recommender systemsrdquoACM Transactions on Information Systems vol 22 no 1pp 5ndash53 2004

[58] P Lops M De Gemmis and G Semeraro ldquoContent-basedrecommender systems state of the art and trendsrdquo in Rec-ommender Systems Handbook pp 73ndash105 Springer BerlinHeidelberg Germany 2011

[59] Z Lu H Wang N Mamoulis W Tu and D W CheungldquoPersonalized location recommendation by aggregatingmultiple recommenders in diversityrdquo GeoInformatica vol 21no 3 pp 459ndash484 2017

[60] R Burke ldquoHybrid recommender systems survey and ex-perimentsrdquo User Modeling and User-Adapted Interactionvol 12 no 4 pp 331ndash370 2002

[61] L Baltrunas and X Amatriain ldquoTowards time-dependentrecommendation based on implicit feedbackrdquo in Pro-ceedings of Workshop on Context-Aware Recommender Sys-tems (CARSrsquo09) New York NY USA October 2009

[62] X Wang Y L Zhao L Nie et al ldquoSemantic-based locationrecommendation with multimodal venue semanticsrdquo IEEETransactions on Multimedia vol 17 no 3 pp 409ndash4192015

[63] Q Yuan G Cong and A Sun ldquoGraph-based point-of-interest recommendation with geographical and temporalinfluencesrdquo in Proceedings of 23rd ACM International Con-ference on Information and Knowledge Managementpp 659ndash668 Shanghai China November 2014

[64] Y Mao P Yin W Lee and D Lee ldquoExploiting geographicalinfluence for collaborative point-of-interest recommenda-tionrdquo in Proceedings of 34th ACM SI- GIR InternationalConference on Research and Development in InformationRetrieval pp 325ndash334 Beijing China July 2011

[65] S K Gorakala and M Usuelli Building a RecommendationSystem with R Packt Publishing Ltd Birmingham UK 2015

[66] Y Zheng L Zhang X Xie andW Y Ma ldquoMining interestinglocations and travel sequences from GPS trajectoriesrdquo inProceedings of 18th International Conference on World WideWeb pp 791ndash800 Madrid Spain April 2009

[67] G Shani and A Gunawardana ldquoEvaluating recommendationsystemsrdquo in Recommender Systems Handbook pp 257ndash297Springer Berlin Heidelberg Germany 2011

[68] J A Konstan B N Miller D Maltz J L HerlockerL R Gordon and J Riedl ldquoGroupLens applying collabo-rative filtering to Usenet newsrdquo Communications of the ACMvol 40 no 3 pp 77ndash87 1997

[69] M Deshpande and G Karypis ldquoItem-based top-n recom-mendation algorithmsrdquo ACM Transactions on InformationSystems vol 22 no 1 pp 143ndash177 2004

[70] Y Koren R Bell and C Volinsky ldquoMatrix factorizationtechniques for recommender systemsrdquo Computer vol 42no 8 pp 30ndash37 2009

[71] Y Koren ldquoFactorization meets the neighborhood a multi-faceted collaborative filtering modelrdquo in Proceedings of 14thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 426ndash434 Las Vegas NV USAAugust 2008

Mobile Information Systems 13

Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems

Hindawiwwwhindawicom

Volume 2018

International Journal of

ReconfigurableComputing

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawiwwwhindawicom Volume 2018

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational Intelligence and Neuroscience

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018

Human-ComputerInteraction

Advances in

Hindawiwwwhindawicom Volume 2018

Scientic Programming

Submit your manuscripts atwwwhindawicom

42 User Study for Contextually Personalized HybRecSysAs was indicated in the literature user studies are veryhelpful to understand whether the recommendations areliked by the users and to collect more detailed data about therecommendation system [45 57 67] Although conductinga user study is difficult time-consuming and costly it issuggested to apply them after the offline experiments inorder to validate the results of the offline experiments [67]

43 Steps of the User Study For this study a user study wasconducted on the users in our dataset For this purpose theTwitter account of each user in our dataset was checked tolearn whether their profiles allow to receive direct messagesOut of 1101 users 195 accounts were open for directmessage )ose users were invited to attend our user studyby direct message and a small incentive (a movie ticket) waspromised if they attend this evaluation

Twenty-four users replied to themessage and accepted toattend the study After that the user evaluation occurred inthe following steps

(1) )e algorithm predicted the ratings of that user to allvenues except the venues they visited

(2) Among the results the algorithm recommendedthree top-rated venues for each 24 users )ese userswere asked via Twitter message whether any of therecommended items attracted their attention

(3) Twenty-one users replied to the recommendationsOnly one of them said that none of the venues wassuitable for them Others were interested in at leastone venue among the recommended ones

(4) A survey which also includes the Foursquare links ofrecommended venues was prepared according to theparticipantsrsquo choices and sent to them

(5) )e users were asked to fill out the survey after theyvisit that recommended venue or after examiningthe Foursquare page of the venues

44 Survey Questions of the User Study )ere are 12 ques-tions in the survey )e first question was asked to un-derstand the appreciation of the participants to the

recommended venues It was asked in 5-point Likert scale (1-Not Like At All 2-Not Like 3-Not Sure 4-Like 5-Like VeryMuch)

)e following four questions were asked to measure theappropriateness of the category price popularity and thelocation of the recommended venue for the participant)eywere also asked using a 5-point Likert scale (1-Not Ap-propriate At All 2-Not Appropriate 3-Not Sure 4-Ap-propriate 5-Very Appropriate)

)e following four questions were asked to understandin which contextual circumstances the user would prefer therecommended venue )ese questions were asked as fixedsum scale questions )e participants were asked to dis-tribute a hundred points to the categories of a contextualvariable according to the tendency of the user to visit thatvenue in these categories

)e last two questions were demographic questions)ey were asked to learn the age gender and the educationlevel of the participants

45 Results of the User Study Participantsrsquo average age is 30and the age range varies from 19 to 38 )ere are eightwomen and 12 men in the dataset 12 of the participantsgraduated from high school six of them have a bachelorrsquosdegree and two of them have a masterrsquos degree Table 12presents the answers of the participants to the first fivequestions of the survey

Ninety percent of the participants liked the recom-mended venues Ten percent of them were indecisive aboutwhether they like the recommended venue Ninety percentof the participants thought that the price class of the rec-ommended venue was suitable for them Forty percent of theparticipants thought that the category of recommendedvenues were very appropriate for them while 20 of themthought that the categories were appropriate and theremaining are indecisive Seventy percent of the participantsthought that the popularity class of recommended venueswas appropriate for them and 10 of them thought that thepopularity classes of the venues were very appropriate Onthe other hand 10 were indecisive while another 10thought that the popularity classes of the venues were notappropriate Moreover the address of the recommended

05

04

03

02

01

0User-based

KNNItem-based

KNNBiased matrixfactorization

SVD++ HybRecSys(neural)

ContextualpersonalizedHybRecSys

PrecisionRecallF1-measure

Figure 2 Graph of precision recall and F-1 measures of the algorithms

8 Mobile Information Systems

venue was given to the participants and they were askedwhether the location was appropriate or not 40 of theparticipants said very appropriate and 30 of them saidappropriate while the other 30 said not appropriate

)e following four questions were asked to understandin which contextual circumstances participants will prefer tovisit the recommended venues Table 13 presents the an-swers of each participant to these questions and the pre-dictions which are calculated from the algorithms for eachuser

)e predicted results were compared with the actualanswers of the participants )e evaluation of the recom-mendation was measured with precision recall and F-1measures Table 14 presents the precision recall and F-1scores for 20 participants Precision value is 0282 recallvalue is 0276 and F1-score is 0279

)e participants made some comments about the venuesvia Twitter messages Some of the participants had been tosome of the restaurants and supported the given recom-mendation with the following expressions

Table 12 Survey answers (Q1ndashQ5)

1 () 2 () 3 () 4 () 5 ()1 Rate your liking 10 80 102 Rate the appropriateness of the price range 10 903 Rate the appropriateness of the category of therestaurant 40 20 40

4 Rate the appropriateness of the popularity class ofthe restaurant 10 10 70 10

5 Rate the appropriateness of the location of therestaurant 30 30 40

Table 13 Answers of questions 6-7-8-9 and prediction of the answers

Answers of questions 6-7-8-9 Prediction of the answersM N E Wdays Wend Hot Cold Sunny Rainy Snowy M N E Wdays Wend Hot Cold Sunny Rainy Snowy

1 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 0002 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 0003 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 0004 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 0005 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 0006 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 0007 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 0008 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 0009 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02510 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 00011 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 00012 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 00013 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 00014 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 00015 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 00016 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 00017 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 00018 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 00019 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02520 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 000

Table 14 Precision recall and F-1 measure for user study

Approaches Precision Recall F-1 scoreUser-based KNN 01220 00823 00983Item-based KNN 01154 01107 01130Biased matrix factorization 00893 01200 01031SVD++ 00702 00976 00816HybRecSys 01667 01460 01493Contextually personalized HybRecSys 018 045 025User study results of contextually personalizedHybRecSys 0282 0276 0279

Mobile Information Systems 9

(i) Participant 1 (Gender Male Age 37 EducationHigh School)

(a) ldquoI always prefer going these two restaurants thatyou have recommendedrdquo

(ii) Participant 2 (Gender Male Age 38 EducationHigh School)

(a) ldquoI have already been to one of the restaurantsrdquo

(iii) Participant 3 (Gender Female Age 19 EducationHigh School)

(a) ldquoI have visited one of the restaurants beforerdquo

(iv) Participant 4 (Gender Female Age 38 EducationBachelorrsquos Degree)

(a) ldquoI have already visited venue1 and venue2rdquo

)ese statements demonstrate the ability of the rec-ommender systemrsquos accurate prediction Even participantsthat had not visited the recommended venues before statedthat they like the recommended venues )is result revealsthat the developed system has the diversity novelty andserendipity dimensions

5 Conclusion

In this study a contextually personalized hybrid recom-mendation model was proposed)is model integrates user-based and item-based collaborative filtering content-basedfiltering together with contextual information in order to getrid of the drawbacks of each approach Different datasources were used to collect the data Visiting history of userswas collected from Twitter Venue characteristics (distancecategory popularity and price classification) were collectedfrom Foursquare and contextual information (weatherseason date and time of visits) related to each visit werecollected from Weather Underground website For content-based filtering the variables distance category popularityand price classification that have not been used before in onealgorithm were used in order to determine content-baseduser similarity Weather conditions season date and timeof each visit were used cumulatively as the properties ofvenues and contextual similarities of venues were utilized)e artificial neural network algorithm was applied to de-termine the weights of each algorithm Ratings coming fromdifferent algorithms (user-based CF item-based CFcontent-based filtering and rating calculated from thecontextual similarities of venues) were used as the predictorsof the actual rating Final ratings were calculated by mul-tiplying the weights retrieved from neural network and theratings from different algorithms In addition in order tomake a more accurate recommendation and make therecommender system contextually personalized our systemcalculates the probability of visiting a venue in a specificcontextual category for each user-venue pair)e decision ofrecommendation of a venue to the specific user is madeaccording to two rules If the calculated rating is greater thanthe average rating of the user and the total contextual score isgreater than two then that venue will be recommended to

that user under the specific contextual circumstances )eHybrid system prunes the disadvantages of each approachthat may occur when they are used separately

Contextually Personalized HybRecSys was compared touser-based and item-based KNN biased matrix factorizationand SVD++ )ese algorithms are evaluated according to themetrics which are used for ranking prediction (precisionrecall and F-1 measure) Training and test datasets werecreated using K-fold cross-validation (K 10) techniqueResults show that contextually personalized HybRecSysoutperforms existing four algorithms according to eachevaluation metric Contextually Personalized HybRecSys ef-fectively overcomes the data sparsity problem using venuecategory popularity and price to model user preferences

In addition recommending very similar venues toprevious visits causes overspecialization )is problem isalso reduced by considering the user preferences with dif-ferent aspects and not just being stuck on only the venuecharacteristics Hence the quality of the recommendation isimproved and the recommender system gained diversitynovelty and serendipity dimensions

)e algorithm partially solves the cold start issue whichcan be caused by both a new user and a new item Even ifa new user rates only one venue the algorithm understandsthe user preferences from the characteristics of the venue(category price and popularity) Moreover the algorithmfigures out the contextual circumstances under which theuser prefers that the specific venue )erefore by looking atthese characteristics the algorithmmay recommend a venueto a new user

)e most important feature that distinguishes the de-veloped algorithm from others is the contextual personal-ization Contextually personalized HybRecSys solvesthis issue and recommends a right venue under the rightconditions

At the beginning the collected data size was large butafter the filtration it became moderately small On theother hand this small dataset generated significant resultsthat may be improved with larger datasets As a futurestudy it is planned to apply contextually personalizedHybRecSys on larger datasets Venue opening and closinghours can be checked for better results In order to solvethe cold start problem entirely the system may recom-mend a venue to a new user by looking at the contextualcircumstances and recommend the most preferred venuebased on these contextual circumstances In additioneven if a new venue is added to the system it can berecommended by looking at the content-related charac-teristics )e online evaluation was conducted with only20 people which is a relatively small sample for obtainingstatistically precise and coherent results )erefore moreusers should be reached for future studies Finally thisalgorithm can be embedded into a mobile application andthis mobile application can be marketed in mobile ap-plication stores )erefore its performance can be mea-sured via online experiments As users download and usethe application they will score the recommendation sothat online feedback can be taken and the algorithm can beimproved

10 Mobile Information Systems

Data Availability

)e Twitter and Foursquare data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

)e authors declare that they have no conflicts of interest

Acknowledgments

)is study was funded by Bogazici University ScientificResearch Fund (grant no 11463)

References

[1] A Bozanta and B Kutlu ldquoHybRecSys content-based con-textual hybrid venue recommender systemrdquo Journal of In-formation Science pp 1ndash15 2018

[2] M Sattari I H Toroslu P Karagoz P Symeonidis andY Manolopoulos ldquoExtended feature combination model forrecommendations in location-based mobile servicesrdquoKnowledge and Information Systems vol 44 no 3 pp 629ndash661 2015

[3] H Wang G Li and J Feng ldquoGroup-based personalizedlocation recommendation on social networksrdquo in WebTechnologies and Applications pp 68ndash80 Springer BerlinGermany 2014

[4] H Yin B Cui Y Sun Z Hu and L Chen ldquoLCARS a spatialitem recommender systemrdquo ACM Transactions on In-formation Systems (TOIS) vol 32 no 3 pp 1ndash37 2014

[5] M H Park J H Hong and S B Cho ldquoLocation-basedrecommendation system using bayesian userrsquos preferencemodel in mobile devicesrdquo in Ubiquitous Intelligence andComputing pp 1130ndash1139 Springer Berlin HeidelbergGermany 2007

[6] K Aihara H Koshiba and H Takeda ldquoBehavioral cost-basedrecommendation model for wanderers in townrdquo in Human-Computer Interaction Towards Mobile and IntelligentInteraction Environments pp 271ndash279 Springer BerlinHeidelberg Germany 2011

[7] Z Yu Y Feng H Xu and X Zhou ldquoRecommending travelpackages based on mobile crowdsourced datardquo IEEE Com-munications Magazine vol 52 no 8 pp 56ndash62 2014

[8] H Yin B Cui L Chen Z Hu and C Zhang ldquoModelinglocation-based user rating profiles for personalized recom-mendationrdquo ACM Transactions on Knowledge Discovery fromData vol 9 no 3 pp 1ndash41 2015

[9] D Zhou S M Rahimi and X Wang ldquoSimilarity-basedprobabilistic category-based location recommendation uti-lizing temporal and geographical influencerdquo InternationalJournal of Data Science and Analytics vol 1 no 2 pp 111ndash121 2016

[10] H Zhu E Chen H Xiong K Yu H Cao and J TianldquoMining mobile user preferences for personalized context-aware recommendationrdquo ACM Transactions on IntelligentSystems and Technology vol 5 no 4 pp 1ndash27 2015

[11] Q Fang C Xu M S Hossain and M G Stcaplrs ldquoA spatial-temporal context-aware personalized location recommen-dation systemrdquo ACM Transactions on Intelligent systems andTechnology vol 7 no 4 pp 1ndash30 2016

[12] Q Yuan G Cong K Zhao Z Ma and A Sun ldquoWho wherewhen and what a nonparametric bayesian approach to

context-aware recommendation and search for twitter usersrdquoACM Transactions on Information Systems vol 33 no 1pp 1ndash31 2015

[13] A Gupta and K Singh ldquoLocation based personalized res-taurant recommendation system for mobile environmentsrdquoin Proceedings of International Conference on Advances inComputing Communications and Informatics (ICACCI)pp 507ndash511 IEEE Mysore India August 2013

[14] J Bao Y Zheng and M F Mokbel ldquoLocation-based andpreference-aware recommendation using sparse geo-socialnetworking datardquo in Proceedings of 20th International Con-ference on Advances in Geographic Information Systemspp 199ndash208 Redondo Beach CA USA November 2012

[15] H Yin Y Sun B Cui Z Hu and L Chen ldquoLcars a location-content-aware recommender systemrdquo in Proceedings of 19thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 221ndash229 Chicago IL USAAugust 2013

[16] M H Kuo L C Chen and C W Liang ldquoBuilding andevaluating a location-based service recommendation systemwith a preference adjustment mechanismrdquo Expert Systemswith Applications vol 36 no 2 pp 3543ndash3554 2009

[17] K Shimada H Uehara and T Endo ldquoA comparative study ofpotential-of-interest days on a sightseeing spot recom-menderrdquo in Proceedings of IIAI 3rd International Conferenceon Advanced Applied Informatics (IIAIAAI) pp 555ndash560Kitakyushu Japan August 2014

[18] L Ahmedi K Rrmoku K Sylejmani and D Shabani ldquoAbimodal social network analysis to recommend points ofinterest to touristsrdquo Social Network Analysis and Miningvol 7 no 1 p 14 2017

[19] H Gao J Tang X Hu and H Liu ldquoContent-aware point ofinterest recommendation on location-based social networksrdquoin Proceedings of the AAAI pp 1721ndash1727 Austin TexasUSA January 2015

[20] Z Zheng Y Chen S Chen L Sun and D Chen ldquoLocation-aware POI recommendation for indoor space by exploitingWiFi logsrdquo Mobile Information Systems vol 2017 Article ID9601404 16 pages 2017

[21] W X Zhao N Zhou A Sun J R Wen J Han andE Y Chang ldquoA time-aware trajectory embedding model fornext-location recommendationrdquo Knowledge and InformationSystems vol 56 no 3 pp 559ndash579 2017

[22] R Srivastava S Hingmire G K Palshikar S Chaurasia andD A Csrs ldquoA context and sequence aware recommendationsystemrdquo in Proceedings of 8th Annual Meeting of the Forum onInformation Retrieval Evaluation pp 8ndash15 Kolkata IndiaDecember 2016

[23] V Subramaniyaswamy V Vijayakumar R Logesh andV Indragandhi ldquoIntelligent travel recommendation systemby mining attributes from community contributed photosrdquoProcedia Computer Science vol 50 pp 447ndash455 2015

[24] L Cao J Luo A C Gallagher X Jin J Han and T S HuangldquoA worldwide tourism recommendation system based ongeotaggedweb photosrdquo in Proceedings of ICASSP pp 2274ndash2277 Dallas Texas USA March 2010

[25] L Guo J Shao K L Tan and Y Y Wheretogo ldquoPersonalizedtravel recommendation for individuals and groupsrdquo in Pro-ceedings of IEEE 15th International Conference onMobile DataManagement vol 1 pp 49ndash58 Brisbane Australia July 2014

[26] I Memon L Chen A Majid M Lv I Hussain and G ChenldquoTravel recommendation using geo-tagged photos in socialmedia for touristrdquoWireless Personal Communications vol 80no 4 pp 1347ndash1362 2015

Mobile Information Systems 11

[27] Z Xu L Chen and G Chen ldquoTopic based context-awaretravel recommendation method exploiting geotaggedphotosrdquo Neurocomputing vol 155 pp 99ndash107 2015

[28] B Dhake S S Lomte R A Auti Y R Nagargoje and B PatilldquoLARS an efficient and scalable location-awarerdquo In-ternational Journal of Scientific Research And Educationvol 2 p 11 2014

[29] M Sarwat J J Levandoski A Eldawy and M F MokbelldquoLARSlowast an efficient and scalable location-aware recom-mender systemrdquo IEEE Transactions on Knowledge and DataEngineering vol 26 no 6 pp 1384ndash1399 2014

[30] P V Krishna S Misra D Joshi andM S Obaidat ldquoLearningautomata based sentiment analysis for recommender systemon cloudrdquo in Proceedings of International Conference onComputer Information and Telecommunication Systems(CITS) pp 1ndash5 IEEE Athens Greece May 2013

[31] M Mordacchini A Passarella M Conti et al ldquoCrowd-sourcing through cognitive opportunistic networksrdquo ACMTransactions on Autonomous and Adaptive Systems vol 10no 2 pp 1ndash29 2015

[32] Y Zheng L Zhang Z Ma X Xie and W Y Ma ldquoRec-ommending friends and locations based on individual loca-tion historyrdquo ACM Transactions on the Web vol 5 no 1pp 1ndash44 2011

[33] VW Zheng Y Zheng X Xie and Q Yang ldquoTowards mobileintelligence learning from GPS history data for collaborativerecommendationrdquo Artificial Intelligence vol 184ndash185pp 17ndash37 2012

[34] M Korakakis E Spyrou P Mylonas and S J PerantonisldquoExploiting social media information toward a context-awarerecommendation systemrdquo Social Network Analysis andMining vol 7 no 1 p 42 2017

[35] Q Li Y Zheng X Xie Y Chen W Liu and W Y MaldquoMining user similarity based on location historyrdquo in Pro-ceedings of 16th ACM SIGSPATIAL International Conferenceon Advances in Geographic Information Systems p 34 ACMNew York NY USA November 2008

[36] M Ye P Yin and W C Lee ldquoLocation recommendation forlocation-based social networksrdquo in Proceedings of 18th SIG-SPATIAL International Conference on Advances in GeographicInformation Systems pp 458ndash461 ACM San Jose CA USANovember 2010

[37] T Hasegawa and T Hayashi ldquoCollaborative filtering basedspot recommendation seamlessly available in home and awayareasrdquo in Proceedings of the IEEEACIS 12th InternationalConference on Computer and Information Science (ICIS)pp 547-548 IEEE Niigata Japan June 2013

[38] J J Levandoski M Sarwat A Eldawy and M F MokbelldquoLars a location-aware recommender systemrdquo in Proceedingsof IEEE 28th International Conference on Data Engineering(ICDE) pp 450ndash461 IEEE Arlington VA USA April 2012

[39] Y Takeuchi and M Sugimoto ldquoCityvoyager an outdoorrecommendation system based on user location historyrdquo inUbiquitous Intelligence and Computing pp 625ndash636Springer-Verlag Berlin Germany 2006

[40] S Knoch A Chapko A Emrich D Werth and P Loos ldquoAcontext-aware running route recommender learning fromuser histories using artificial neural networksrdquo in Proceedingsof 23rd International Workshop on Database and ExpertSystems Applications (DEXA) pp 106ndash110 IEEE ViennaAustria September 2012

[41] M Saraee S Khan and S Yamaner ldquoData mining approachto implement a recommendation system for electronic tourguidesrdquo in Proceedings of 2005 International Conference on

E-Business Enterprise Information Systems E-Governmentand Outsourcing EEE 2005 pp 215ndash218 CSREA PressLas Vegas NV USA June 2005

[42] R Katarya and O P Verma ldquoRestaurant recommendersystem based on psychographic and demographic factors inmobile environmentrdquo in Proceedings of International Con-ference on Green Computing and Internet of Dings (ICG-CIoT) pp 907ndash912 Uttar Pradesh India October 2015

[43] R Katarya and O P Verma ldquoRecent developments in af-fective recommender systemsrdquo Physica A Statistical Me-chanics and its Applications vol 461 pp 182ndash190 2016

[44] R Katarya ldquoA systematic review of group recommendersystems techniquesrdquo in Proceedings of International Confer-ence on Intelligent Sustainable Systems (ICISS) pp 425ndash428Tirupur India December 2017

[45] R Katarya O P Verma and I Jain ldquoUser behaviour analysisin context-aware recommender system using hybrid filteringapproachrdquo in Proceedings of 4th International Conference onComputer and Communication Technology (ICCCT)pp 222ndash227 Allahabad India September 2013

[46] R Katarya M Ranjan and O P Verma ldquoLocation basedrecommender system using enhanced random walk modelrdquoin Proceedings of Fourth International Conference on ParallelDistributed and Grid Computing (PDGC) pp 33ndash37Waknaghat India December 2016

[47] K Waga A Tabarcea and P Franti ldquoContext aware rec-ommendation of location-based datardquo in Proceedings of15th International Conference on System Deory Controland Computing (ICSTCC) pp 1ndash6 IEEE Sinaia RomaniaOctober 2011

[48] A Majid L Chen G Chen H T Mirza I Hussain andJ Woodward ldquoA context-aware personalized travel recom-mendation system based on geotagged social media dataminingrdquo International Journal of Geographical InformationScience vol 27 no 4 pp 662ndash684 2013

[49] Q Yuan G Cong Z Ma A Sun and N M)almann ldquoTime-aware point-of-interest recommendationrdquo in Proceedings of36th International ACM SIGIR Conference on Research andDevelopment in Information Retrieval pp 363ndash372 ACMDublin Ireland July 2013

[50] R Baral and T Li ldquoMaps a multi aspect personalized poirecommender systemrdquo in Proceedings of 10th ACM Confer-ence on Recommender Systems pp 281ndash284 ACM BostonMA USA September 2016

[51] P Hiesel M Braunhofer and W Worndl ldquoLearning thepopularity of items for mobile tourist guidesrdquo in Proceedingsof RecTour Cambridge MA USA September 2016

[52] L Yao Q Z Sheng X Wang W E Zhang and Y QinldquoCollaborative location recommendation by integratingmulti-dimensional contextual informationrdquo ACM Trans-actions on Internet Technology vol 18 no 3 pp 1ndash24 2018

[53] C Trattner A Oberegger L Eberhard D Parra andL Marinho ldquoUnderstanding the impact of weather for POIrecommendationsrdquo in Proceedings of RecTour CambridgeMA USA September 2016

[54] Y Zheng R Burke and B Mobasher ldquoDifferential contextrelaxation for context-aware travel recommendationrdquo inProceedings of International Conference on Electronic Com-merce and Web Technologies pp 88ndash99 Springer ViennaAustria September 2012

[55] M J Barranco J M Noguera J Castro and L Martinez ldquoAcontext-aware mobile recommender system based on locationand trajectoryrdquo in Management Intelligent Systems pp 153ndash162 Springer Berlin Heidelberg Germany 2012

12 Mobile Information Systems

[56] N S Savage M Baranski N E Chavez and T Hollerer ldquoIrsquomfeeling loco a location based context aware recommendationsystemrdquo in Advances in Location-Based Services pp 37ndash54Springer Berlin Heidelberg Germany 2012

[57] J L Herlocker J A Konstan L G Terveen and J T RiedlldquoEvaluating collaborative filtering recommender systemsrdquoACM Transactions on Information Systems vol 22 no 1pp 5ndash53 2004

[58] P Lops M De Gemmis and G Semeraro ldquoContent-basedrecommender systems state of the art and trendsrdquo in Rec-ommender Systems Handbook pp 73ndash105 Springer BerlinHeidelberg Germany 2011

[59] Z Lu H Wang N Mamoulis W Tu and D W CheungldquoPersonalized location recommendation by aggregatingmultiple recommenders in diversityrdquo GeoInformatica vol 21no 3 pp 459ndash484 2017

[60] R Burke ldquoHybrid recommender systems survey and ex-perimentsrdquo User Modeling and User-Adapted Interactionvol 12 no 4 pp 331ndash370 2002

[61] L Baltrunas and X Amatriain ldquoTowards time-dependentrecommendation based on implicit feedbackrdquo in Pro-ceedings of Workshop on Context-Aware Recommender Sys-tems (CARSrsquo09) New York NY USA October 2009

[62] X Wang Y L Zhao L Nie et al ldquoSemantic-based locationrecommendation with multimodal venue semanticsrdquo IEEETransactions on Multimedia vol 17 no 3 pp 409ndash4192015

[63] Q Yuan G Cong and A Sun ldquoGraph-based point-of-interest recommendation with geographical and temporalinfluencesrdquo in Proceedings of 23rd ACM International Con-ference on Information and Knowledge Managementpp 659ndash668 Shanghai China November 2014

[64] Y Mao P Yin W Lee and D Lee ldquoExploiting geographicalinfluence for collaborative point-of-interest recommenda-tionrdquo in Proceedings of 34th ACM SI- GIR InternationalConference on Research and Development in InformationRetrieval pp 325ndash334 Beijing China July 2011

[65] S K Gorakala and M Usuelli Building a RecommendationSystem with R Packt Publishing Ltd Birmingham UK 2015

[66] Y Zheng L Zhang X Xie andW Y Ma ldquoMining interestinglocations and travel sequences from GPS trajectoriesrdquo inProceedings of 18th International Conference on World WideWeb pp 791ndash800 Madrid Spain April 2009

[67] G Shani and A Gunawardana ldquoEvaluating recommendationsystemsrdquo in Recommender Systems Handbook pp 257ndash297Springer Berlin Heidelberg Germany 2011

[68] J A Konstan B N Miller D Maltz J L HerlockerL R Gordon and J Riedl ldquoGroupLens applying collabo-rative filtering to Usenet newsrdquo Communications of the ACMvol 40 no 3 pp 77ndash87 1997

[69] M Deshpande and G Karypis ldquoItem-based top-n recom-mendation algorithmsrdquo ACM Transactions on InformationSystems vol 22 no 1 pp 143ndash177 2004

[70] Y Koren R Bell and C Volinsky ldquoMatrix factorizationtechniques for recommender systemsrdquo Computer vol 42no 8 pp 30ndash37 2009

[71] Y Koren ldquoFactorization meets the neighborhood a multi-faceted collaborative filtering modelrdquo in Proceedings of 14thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 426ndash434 Las Vegas NV USAAugust 2008

Mobile Information Systems 13

Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems

Hindawiwwwhindawicom

Volume 2018

International Journal of

ReconfigurableComputing

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawiwwwhindawicom Volume 2018

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational Intelligence and Neuroscience

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018

Human-ComputerInteraction

Advances in

Hindawiwwwhindawicom Volume 2018

Scientic Programming

Submit your manuscripts atwwwhindawicom

venue was given to the participants and they were askedwhether the location was appropriate or not 40 of theparticipants said very appropriate and 30 of them saidappropriate while the other 30 said not appropriate

)e following four questions were asked to understandin which contextual circumstances participants will prefer tovisit the recommended venues Table 13 presents the an-swers of each participant to these questions and the pre-dictions which are calculated from the algorithms for eachuser

)e predicted results were compared with the actualanswers of the participants )e evaluation of the recom-mendation was measured with precision recall and F-1measures Table 14 presents the precision recall and F-1scores for 20 participants Precision value is 0282 recallvalue is 0276 and F1-score is 0279

)e participants made some comments about the venuesvia Twitter messages Some of the participants had been tosome of the restaurants and supported the given recom-mendation with the following expressions

Table 12 Survey answers (Q1ndashQ5)

1 () 2 () 3 () 4 () 5 ()1 Rate your liking 10 80 102 Rate the appropriateness of the price range 10 903 Rate the appropriateness of the category of therestaurant 40 20 40

4 Rate the appropriateness of the popularity class ofthe restaurant 10 10 70 10

5 Rate the appropriateness of the location of therestaurant 30 30 40

Table 13 Answers of questions 6-7-8-9 and prediction of the answers

Answers of questions 6-7-8-9 Prediction of the answersM N E Wdays Wend Hot Cold Sunny Rainy Snowy M N E Wdays Wend Hot Cold Sunny Rainy Snowy

1 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 0002 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 0003 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 0004 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 0005 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 0006 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 0007 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 0008 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 0009 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02510 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 00011 00 00 100 025 075 050 050 070 025 005 000 000 100 100 000 100 000 100 000 00012 00 00 100 050 050 040 060 00 050 050 000 000 100 100 000 100 000 000 100 00013 00 035 065 065 035 070 030 070 030 00 000 000 100 100 000 000 100 032 068 00014 00 020 080 00 0100 060 040 090 010 00 000 054 046 082 018 037 063 061 039 00015 040 025 035 030 070 050 050 070 010 020 000 000 100 100 000 050 050 050 050 00016 030 010 060 040 060 060 040 080 015 005 000 050 050 000 100 000 100 100 000 00017 00 050 050 080 020 080 020 060 020 020 000 000 100 048 052 052 048 052 048 00018 00 080 020 075 025 075 025 060 020 020 000 000 100 048 052 052 048 052 048 00019 00 040 060 030 070 060 040 020 050 030 000 069 031 043 057 050 050 056 019 02520 00 040 060 020 080 085 015 050 025 025 000 000 100 100 000 000 100 000 100 000

Table 14 Precision recall and F-1 measure for user study

Approaches Precision Recall F-1 scoreUser-based KNN 01220 00823 00983Item-based KNN 01154 01107 01130Biased matrix factorization 00893 01200 01031SVD++ 00702 00976 00816HybRecSys 01667 01460 01493Contextually personalized HybRecSys 018 045 025User study results of contextually personalizedHybRecSys 0282 0276 0279

Mobile Information Systems 9

(i) Participant 1 (Gender Male Age 37 EducationHigh School)

(a) ldquoI always prefer going these two restaurants thatyou have recommendedrdquo

(ii) Participant 2 (Gender Male Age 38 EducationHigh School)

(a) ldquoI have already been to one of the restaurantsrdquo

(iii) Participant 3 (Gender Female Age 19 EducationHigh School)

(a) ldquoI have visited one of the restaurants beforerdquo

(iv) Participant 4 (Gender Female Age 38 EducationBachelorrsquos Degree)

(a) ldquoI have already visited venue1 and venue2rdquo

)ese statements demonstrate the ability of the rec-ommender systemrsquos accurate prediction Even participantsthat had not visited the recommended venues before statedthat they like the recommended venues )is result revealsthat the developed system has the diversity novelty andserendipity dimensions

5 Conclusion

In this study a contextually personalized hybrid recom-mendation model was proposed)is model integrates user-based and item-based collaborative filtering content-basedfiltering together with contextual information in order to getrid of the drawbacks of each approach Different datasources were used to collect the data Visiting history of userswas collected from Twitter Venue characteristics (distancecategory popularity and price classification) were collectedfrom Foursquare and contextual information (weatherseason date and time of visits) related to each visit werecollected from Weather Underground website For content-based filtering the variables distance category popularityand price classification that have not been used before in onealgorithm were used in order to determine content-baseduser similarity Weather conditions season date and timeof each visit were used cumulatively as the properties ofvenues and contextual similarities of venues were utilized)e artificial neural network algorithm was applied to de-termine the weights of each algorithm Ratings coming fromdifferent algorithms (user-based CF item-based CFcontent-based filtering and rating calculated from thecontextual similarities of venues) were used as the predictorsof the actual rating Final ratings were calculated by mul-tiplying the weights retrieved from neural network and theratings from different algorithms In addition in order tomake a more accurate recommendation and make therecommender system contextually personalized our systemcalculates the probability of visiting a venue in a specificcontextual category for each user-venue pair)e decision ofrecommendation of a venue to the specific user is madeaccording to two rules If the calculated rating is greater thanthe average rating of the user and the total contextual score isgreater than two then that venue will be recommended to

that user under the specific contextual circumstances )eHybrid system prunes the disadvantages of each approachthat may occur when they are used separately

Contextually Personalized HybRecSys was compared touser-based and item-based KNN biased matrix factorizationand SVD++ )ese algorithms are evaluated according to themetrics which are used for ranking prediction (precisionrecall and F-1 measure) Training and test datasets werecreated using K-fold cross-validation (K 10) techniqueResults show that contextually personalized HybRecSysoutperforms existing four algorithms according to eachevaluation metric Contextually Personalized HybRecSys ef-fectively overcomes the data sparsity problem using venuecategory popularity and price to model user preferences

In addition recommending very similar venues toprevious visits causes overspecialization )is problem isalso reduced by considering the user preferences with dif-ferent aspects and not just being stuck on only the venuecharacteristics Hence the quality of the recommendation isimproved and the recommender system gained diversitynovelty and serendipity dimensions

)e algorithm partially solves the cold start issue whichcan be caused by both a new user and a new item Even ifa new user rates only one venue the algorithm understandsthe user preferences from the characteristics of the venue(category price and popularity) Moreover the algorithmfigures out the contextual circumstances under which theuser prefers that the specific venue )erefore by looking atthese characteristics the algorithmmay recommend a venueto a new user

)e most important feature that distinguishes the de-veloped algorithm from others is the contextual personal-ization Contextually personalized HybRecSys solvesthis issue and recommends a right venue under the rightconditions

At the beginning the collected data size was large butafter the filtration it became moderately small On theother hand this small dataset generated significant resultsthat may be improved with larger datasets As a futurestudy it is planned to apply contextually personalizedHybRecSys on larger datasets Venue opening and closinghours can be checked for better results In order to solvethe cold start problem entirely the system may recom-mend a venue to a new user by looking at the contextualcircumstances and recommend the most preferred venuebased on these contextual circumstances In additioneven if a new venue is added to the system it can berecommended by looking at the content-related charac-teristics )e online evaluation was conducted with only20 people which is a relatively small sample for obtainingstatistically precise and coherent results )erefore moreusers should be reached for future studies Finally thisalgorithm can be embedded into a mobile application andthis mobile application can be marketed in mobile ap-plication stores )erefore its performance can be mea-sured via online experiments As users download and usethe application they will score the recommendation sothat online feedback can be taken and the algorithm can beimproved

10 Mobile Information Systems

Data Availability

)e Twitter and Foursquare data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

)e authors declare that they have no conflicts of interest

Acknowledgments

)is study was funded by Bogazici University ScientificResearch Fund (grant no 11463)

References

[1] A Bozanta and B Kutlu ldquoHybRecSys content-based con-textual hybrid venue recommender systemrdquo Journal of In-formation Science pp 1ndash15 2018

[2] M Sattari I H Toroslu P Karagoz P Symeonidis andY Manolopoulos ldquoExtended feature combination model forrecommendations in location-based mobile servicesrdquoKnowledge and Information Systems vol 44 no 3 pp 629ndash661 2015

[3] H Wang G Li and J Feng ldquoGroup-based personalizedlocation recommendation on social networksrdquo in WebTechnologies and Applications pp 68ndash80 Springer BerlinGermany 2014

[4] H Yin B Cui Y Sun Z Hu and L Chen ldquoLCARS a spatialitem recommender systemrdquo ACM Transactions on In-formation Systems (TOIS) vol 32 no 3 pp 1ndash37 2014

[5] M H Park J H Hong and S B Cho ldquoLocation-basedrecommendation system using bayesian userrsquos preferencemodel in mobile devicesrdquo in Ubiquitous Intelligence andComputing pp 1130ndash1139 Springer Berlin HeidelbergGermany 2007

[6] K Aihara H Koshiba and H Takeda ldquoBehavioral cost-basedrecommendation model for wanderers in townrdquo in Human-Computer Interaction Towards Mobile and IntelligentInteraction Environments pp 271ndash279 Springer BerlinHeidelberg Germany 2011

[7] Z Yu Y Feng H Xu and X Zhou ldquoRecommending travelpackages based on mobile crowdsourced datardquo IEEE Com-munications Magazine vol 52 no 8 pp 56ndash62 2014

[8] H Yin B Cui L Chen Z Hu and C Zhang ldquoModelinglocation-based user rating profiles for personalized recom-mendationrdquo ACM Transactions on Knowledge Discovery fromData vol 9 no 3 pp 1ndash41 2015

[9] D Zhou S M Rahimi and X Wang ldquoSimilarity-basedprobabilistic category-based location recommendation uti-lizing temporal and geographical influencerdquo InternationalJournal of Data Science and Analytics vol 1 no 2 pp 111ndash121 2016

[10] H Zhu E Chen H Xiong K Yu H Cao and J TianldquoMining mobile user preferences for personalized context-aware recommendationrdquo ACM Transactions on IntelligentSystems and Technology vol 5 no 4 pp 1ndash27 2015

[11] Q Fang C Xu M S Hossain and M G Stcaplrs ldquoA spatial-temporal context-aware personalized location recommen-dation systemrdquo ACM Transactions on Intelligent systems andTechnology vol 7 no 4 pp 1ndash30 2016

[12] Q Yuan G Cong K Zhao Z Ma and A Sun ldquoWho wherewhen and what a nonparametric bayesian approach to

context-aware recommendation and search for twitter usersrdquoACM Transactions on Information Systems vol 33 no 1pp 1ndash31 2015

[13] A Gupta and K Singh ldquoLocation based personalized res-taurant recommendation system for mobile environmentsrdquoin Proceedings of International Conference on Advances inComputing Communications and Informatics (ICACCI)pp 507ndash511 IEEE Mysore India August 2013

[14] J Bao Y Zheng and M F Mokbel ldquoLocation-based andpreference-aware recommendation using sparse geo-socialnetworking datardquo in Proceedings of 20th International Con-ference on Advances in Geographic Information Systemspp 199ndash208 Redondo Beach CA USA November 2012

[15] H Yin Y Sun B Cui Z Hu and L Chen ldquoLcars a location-content-aware recommender systemrdquo in Proceedings of 19thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 221ndash229 Chicago IL USAAugust 2013

[16] M H Kuo L C Chen and C W Liang ldquoBuilding andevaluating a location-based service recommendation systemwith a preference adjustment mechanismrdquo Expert Systemswith Applications vol 36 no 2 pp 3543ndash3554 2009

[17] K Shimada H Uehara and T Endo ldquoA comparative study ofpotential-of-interest days on a sightseeing spot recom-menderrdquo in Proceedings of IIAI 3rd International Conferenceon Advanced Applied Informatics (IIAIAAI) pp 555ndash560Kitakyushu Japan August 2014

[18] L Ahmedi K Rrmoku K Sylejmani and D Shabani ldquoAbimodal social network analysis to recommend points ofinterest to touristsrdquo Social Network Analysis and Miningvol 7 no 1 p 14 2017

[19] H Gao J Tang X Hu and H Liu ldquoContent-aware point ofinterest recommendation on location-based social networksrdquoin Proceedings of the AAAI pp 1721ndash1727 Austin TexasUSA January 2015

[20] Z Zheng Y Chen S Chen L Sun and D Chen ldquoLocation-aware POI recommendation for indoor space by exploitingWiFi logsrdquo Mobile Information Systems vol 2017 Article ID9601404 16 pages 2017

[21] W X Zhao N Zhou A Sun J R Wen J Han andE Y Chang ldquoA time-aware trajectory embedding model fornext-location recommendationrdquo Knowledge and InformationSystems vol 56 no 3 pp 559ndash579 2017

[22] R Srivastava S Hingmire G K Palshikar S Chaurasia andD A Csrs ldquoA context and sequence aware recommendationsystemrdquo in Proceedings of 8th Annual Meeting of the Forum onInformation Retrieval Evaluation pp 8ndash15 Kolkata IndiaDecember 2016

[23] V Subramaniyaswamy V Vijayakumar R Logesh andV Indragandhi ldquoIntelligent travel recommendation systemby mining attributes from community contributed photosrdquoProcedia Computer Science vol 50 pp 447ndash455 2015

[24] L Cao J Luo A C Gallagher X Jin J Han and T S HuangldquoA worldwide tourism recommendation system based ongeotaggedweb photosrdquo in Proceedings of ICASSP pp 2274ndash2277 Dallas Texas USA March 2010

[25] L Guo J Shao K L Tan and Y Y Wheretogo ldquoPersonalizedtravel recommendation for individuals and groupsrdquo in Pro-ceedings of IEEE 15th International Conference onMobile DataManagement vol 1 pp 49ndash58 Brisbane Australia July 2014

[26] I Memon L Chen A Majid M Lv I Hussain and G ChenldquoTravel recommendation using geo-tagged photos in socialmedia for touristrdquoWireless Personal Communications vol 80no 4 pp 1347ndash1362 2015

Mobile Information Systems 11

[27] Z Xu L Chen and G Chen ldquoTopic based context-awaretravel recommendation method exploiting geotaggedphotosrdquo Neurocomputing vol 155 pp 99ndash107 2015

[28] B Dhake S S Lomte R A Auti Y R Nagargoje and B PatilldquoLARS an efficient and scalable location-awarerdquo In-ternational Journal of Scientific Research And Educationvol 2 p 11 2014

[29] M Sarwat J J Levandoski A Eldawy and M F MokbelldquoLARSlowast an efficient and scalable location-aware recom-mender systemrdquo IEEE Transactions on Knowledge and DataEngineering vol 26 no 6 pp 1384ndash1399 2014

[30] P V Krishna S Misra D Joshi andM S Obaidat ldquoLearningautomata based sentiment analysis for recommender systemon cloudrdquo in Proceedings of International Conference onComputer Information and Telecommunication Systems(CITS) pp 1ndash5 IEEE Athens Greece May 2013

[31] M Mordacchini A Passarella M Conti et al ldquoCrowd-sourcing through cognitive opportunistic networksrdquo ACMTransactions on Autonomous and Adaptive Systems vol 10no 2 pp 1ndash29 2015

[32] Y Zheng L Zhang Z Ma X Xie and W Y Ma ldquoRec-ommending friends and locations based on individual loca-tion historyrdquo ACM Transactions on the Web vol 5 no 1pp 1ndash44 2011

[33] VW Zheng Y Zheng X Xie and Q Yang ldquoTowards mobileintelligence learning from GPS history data for collaborativerecommendationrdquo Artificial Intelligence vol 184ndash185pp 17ndash37 2012

[34] M Korakakis E Spyrou P Mylonas and S J PerantonisldquoExploiting social media information toward a context-awarerecommendation systemrdquo Social Network Analysis andMining vol 7 no 1 p 42 2017

[35] Q Li Y Zheng X Xie Y Chen W Liu and W Y MaldquoMining user similarity based on location historyrdquo in Pro-ceedings of 16th ACM SIGSPATIAL International Conferenceon Advances in Geographic Information Systems p 34 ACMNew York NY USA November 2008

[36] M Ye P Yin and W C Lee ldquoLocation recommendation forlocation-based social networksrdquo in Proceedings of 18th SIG-SPATIAL International Conference on Advances in GeographicInformation Systems pp 458ndash461 ACM San Jose CA USANovember 2010

[37] T Hasegawa and T Hayashi ldquoCollaborative filtering basedspot recommendation seamlessly available in home and awayareasrdquo in Proceedings of the IEEEACIS 12th InternationalConference on Computer and Information Science (ICIS)pp 547-548 IEEE Niigata Japan June 2013

[38] J J Levandoski M Sarwat A Eldawy and M F MokbelldquoLars a location-aware recommender systemrdquo in Proceedingsof IEEE 28th International Conference on Data Engineering(ICDE) pp 450ndash461 IEEE Arlington VA USA April 2012

[39] Y Takeuchi and M Sugimoto ldquoCityvoyager an outdoorrecommendation system based on user location historyrdquo inUbiquitous Intelligence and Computing pp 625ndash636Springer-Verlag Berlin Germany 2006

[40] S Knoch A Chapko A Emrich D Werth and P Loos ldquoAcontext-aware running route recommender learning fromuser histories using artificial neural networksrdquo in Proceedingsof 23rd International Workshop on Database and ExpertSystems Applications (DEXA) pp 106ndash110 IEEE ViennaAustria September 2012

[41] M Saraee S Khan and S Yamaner ldquoData mining approachto implement a recommendation system for electronic tourguidesrdquo in Proceedings of 2005 International Conference on

E-Business Enterprise Information Systems E-Governmentand Outsourcing EEE 2005 pp 215ndash218 CSREA PressLas Vegas NV USA June 2005

[42] R Katarya and O P Verma ldquoRestaurant recommendersystem based on psychographic and demographic factors inmobile environmentrdquo in Proceedings of International Con-ference on Green Computing and Internet of Dings (ICG-CIoT) pp 907ndash912 Uttar Pradesh India October 2015

[43] R Katarya and O P Verma ldquoRecent developments in af-fective recommender systemsrdquo Physica A Statistical Me-chanics and its Applications vol 461 pp 182ndash190 2016

[44] R Katarya ldquoA systematic review of group recommendersystems techniquesrdquo in Proceedings of International Confer-ence on Intelligent Sustainable Systems (ICISS) pp 425ndash428Tirupur India December 2017

[45] R Katarya O P Verma and I Jain ldquoUser behaviour analysisin context-aware recommender system using hybrid filteringapproachrdquo in Proceedings of 4th International Conference onComputer and Communication Technology (ICCCT)pp 222ndash227 Allahabad India September 2013

[46] R Katarya M Ranjan and O P Verma ldquoLocation basedrecommender system using enhanced random walk modelrdquoin Proceedings of Fourth International Conference on ParallelDistributed and Grid Computing (PDGC) pp 33ndash37Waknaghat India December 2016

[47] K Waga A Tabarcea and P Franti ldquoContext aware rec-ommendation of location-based datardquo in Proceedings of15th International Conference on System Deory Controland Computing (ICSTCC) pp 1ndash6 IEEE Sinaia RomaniaOctober 2011

[48] A Majid L Chen G Chen H T Mirza I Hussain andJ Woodward ldquoA context-aware personalized travel recom-mendation system based on geotagged social media dataminingrdquo International Journal of Geographical InformationScience vol 27 no 4 pp 662ndash684 2013

[49] Q Yuan G Cong Z Ma A Sun and N M)almann ldquoTime-aware point-of-interest recommendationrdquo in Proceedings of36th International ACM SIGIR Conference on Research andDevelopment in Information Retrieval pp 363ndash372 ACMDublin Ireland July 2013

[50] R Baral and T Li ldquoMaps a multi aspect personalized poirecommender systemrdquo in Proceedings of 10th ACM Confer-ence on Recommender Systems pp 281ndash284 ACM BostonMA USA September 2016

[51] P Hiesel M Braunhofer and W Worndl ldquoLearning thepopularity of items for mobile tourist guidesrdquo in Proceedingsof RecTour Cambridge MA USA September 2016

[52] L Yao Q Z Sheng X Wang W E Zhang and Y QinldquoCollaborative location recommendation by integratingmulti-dimensional contextual informationrdquo ACM Trans-actions on Internet Technology vol 18 no 3 pp 1ndash24 2018

[53] C Trattner A Oberegger L Eberhard D Parra andL Marinho ldquoUnderstanding the impact of weather for POIrecommendationsrdquo in Proceedings of RecTour CambridgeMA USA September 2016

[54] Y Zheng R Burke and B Mobasher ldquoDifferential contextrelaxation for context-aware travel recommendationrdquo inProceedings of International Conference on Electronic Com-merce and Web Technologies pp 88ndash99 Springer ViennaAustria September 2012

[55] M J Barranco J M Noguera J Castro and L Martinez ldquoAcontext-aware mobile recommender system based on locationand trajectoryrdquo in Management Intelligent Systems pp 153ndash162 Springer Berlin Heidelberg Germany 2012

12 Mobile Information Systems

[56] N S Savage M Baranski N E Chavez and T Hollerer ldquoIrsquomfeeling loco a location based context aware recommendationsystemrdquo in Advances in Location-Based Services pp 37ndash54Springer Berlin Heidelberg Germany 2012

[57] J L Herlocker J A Konstan L G Terveen and J T RiedlldquoEvaluating collaborative filtering recommender systemsrdquoACM Transactions on Information Systems vol 22 no 1pp 5ndash53 2004

[58] P Lops M De Gemmis and G Semeraro ldquoContent-basedrecommender systems state of the art and trendsrdquo in Rec-ommender Systems Handbook pp 73ndash105 Springer BerlinHeidelberg Germany 2011

[59] Z Lu H Wang N Mamoulis W Tu and D W CheungldquoPersonalized location recommendation by aggregatingmultiple recommenders in diversityrdquo GeoInformatica vol 21no 3 pp 459ndash484 2017

[60] R Burke ldquoHybrid recommender systems survey and ex-perimentsrdquo User Modeling and User-Adapted Interactionvol 12 no 4 pp 331ndash370 2002

[61] L Baltrunas and X Amatriain ldquoTowards time-dependentrecommendation based on implicit feedbackrdquo in Pro-ceedings of Workshop on Context-Aware Recommender Sys-tems (CARSrsquo09) New York NY USA October 2009

[62] X Wang Y L Zhao L Nie et al ldquoSemantic-based locationrecommendation with multimodal venue semanticsrdquo IEEETransactions on Multimedia vol 17 no 3 pp 409ndash4192015

[63] Q Yuan G Cong and A Sun ldquoGraph-based point-of-interest recommendation with geographical and temporalinfluencesrdquo in Proceedings of 23rd ACM International Con-ference on Information and Knowledge Managementpp 659ndash668 Shanghai China November 2014

[64] Y Mao P Yin W Lee and D Lee ldquoExploiting geographicalinfluence for collaborative point-of-interest recommenda-tionrdquo in Proceedings of 34th ACM SI- GIR InternationalConference on Research and Development in InformationRetrieval pp 325ndash334 Beijing China July 2011

[65] S K Gorakala and M Usuelli Building a RecommendationSystem with R Packt Publishing Ltd Birmingham UK 2015

[66] Y Zheng L Zhang X Xie andW Y Ma ldquoMining interestinglocations and travel sequences from GPS trajectoriesrdquo inProceedings of 18th International Conference on World WideWeb pp 791ndash800 Madrid Spain April 2009

[67] G Shani and A Gunawardana ldquoEvaluating recommendationsystemsrdquo in Recommender Systems Handbook pp 257ndash297Springer Berlin Heidelberg Germany 2011

[68] J A Konstan B N Miller D Maltz J L HerlockerL R Gordon and J Riedl ldquoGroupLens applying collabo-rative filtering to Usenet newsrdquo Communications of the ACMvol 40 no 3 pp 77ndash87 1997

[69] M Deshpande and G Karypis ldquoItem-based top-n recom-mendation algorithmsrdquo ACM Transactions on InformationSystems vol 22 no 1 pp 143ndash177 2004

[70] Y Koren R Bell and C Volinsky ldquoMatrix factorizationtechniques for recommender systemsrdquo Computer vol 42no 8 pp 30ndash37 2009

[71] Y Koren ldquoFactorization meets the neighborhood a multi-faceted collaborative filtering modelrdquo in Proceedings of 14thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 426ndash434 Las Vegas NV USAAugust 2008

Mobile Information Systems 13

Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems

Hindawiwwwhindawicom

Volume 2018

International Journal of

ReconfigurableComputing

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawiwwwhindawicom Volume 2018

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational Intelligence and Neuroscience

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018

Human-ComputerInteraction

Advances in

Hindawiwwwhindawicom Volume 2018

Scientic Programming

Submit your manuscripts atwwwhindawicom

(i) Participant 1 (Gender Male Age 37 EducationHigh School)

(a) ldquoI always prefer going these two restaurants thatyou have recommendedrdquo

(ii) Participant 2 (Gender Male Age 38 EducationHigh School)

(a) ldquoI have already been to one of the restaurantsrdquo

(iii) Participant 3 (Gender Female Age 19 EducationHigh School)

(a) ldquoI have visited one of the restaurants beforerdquo

(iv) Participant 4 (Gender Female Age 38 EducationBachelorrsquos Degree)

(a) ldquoI have already visited venue1 and venue2rdquo

)ese statements demonstrate the ability of the rec-ommender systemrsquos accurate prediction Even participantsthat had not visited the recommended venues before statedthat they like the recommended venues )is result revealsthat the developed system has the diversity novelty andserendipity dimensions

5 Conclusion

In this study a contextually personalized hybrid recom-mendation model was proposed)is model integrates user-based and item-based collaborative filtering content-basedfiltering together with contextual information in order to getrid of the drawbacks of each approach Different datasources were used to collect the data Visiting history of userswas collected from Twitter Venue characteristics (distancecategory popularity and price classification) were collectedfrom Foursquare and contextual information (weatherseason date and time of visits) related to each visit werecollected from Weather Underground website For content-based filtering the variables distance category popularityand price classification that have not been used before in onealgorithm were used in order to determine content-baseduser similarity Weather conditions season date and timeof each visit were used cumulatively as the properties ofvenues and contextual similarities of venues were utilized)e artificial neural network algorithm was applied to de-termine the weights of each algorithm Ratings coming fromdifferent algorithms (user-based CF item-based CFcontent-based filtering and rating calculated from thecontextual similarities of venues) were used as the predictorsof the actual rating Final ratings were calculated by mul-tiplying the weights retrieved from neural network and theratings from different algorithms In addition in order tomake a more accurate recommendation and make therecommender system contextually personalized our systemcalculates the probability of visiting a venue in a specificcontextual category for each user-venue pair)e decision ofrecommendation of a venue to the specific user is madeaccording to two rules If the calculated rating is greater thanthe average rating of the user and the total contextual score isgreater than two then that venue will be recommended to

that user under the specific contextual circumstances )eHybrid system prunes the disadvantages of each approachthat may occur when they are used separately

Contextually Personalized HybRecSys was compared touser-based and item-based KNN biased matrix factorizationand SVD++ )ese algorithms are evaluated according to themetrics which are used for ranking prediction (precisionrecall and F-1 measure) Training and test datasets werecreated using K-fold cross-validation (K 10) techniqueResults show that contextually personalized HybRecSysoutperforms existing four algorithms according to eachevaluation metric Contextually Personalized HybRecSys ef-fectively overcomes the data sparsity problem using venuecategory popularity and price to model user preferences

In addition recommending very similar venues toprevious visits causes overspecialization )is problem isalso reduced by considering the user preferences with dif-ferent aspects and not just being stuck on only the venuecharacteristics Hence the quality of the recommendation isimproved and the recommender system gained diversitynovelty and serendipity dimensions

)e algorithm partially solves the cold start issue whichcan be caused by both a new user and a new item Even ifa new user rates only one venue the algorithm understandsthe user preferences from the characteristics of the venue(category price and popularity) Moreover the algorithmfigures out the contextual circumstances under which theuser prefers that the specific venue )erefore by looking atthese characteristics the algorithmmay recommend a venueto a new user

)e most important feature that distinguishes the de-veloped algorithm from others is the contextual personal-ization Contextually personalized HybRecSys solvesthis issue and recommends a right venue under the rightconditions

At the beginning the collected data size was large butafter the filtration it became moderately small On theother hand this small dataset generated significant resultsthat may be improved with larger datasets As a futurestudy it is planned to apply contextually personalizedHybRecSys on larger datasets Venue opening and closinghours can be checked for better results In order to solvethe cold start problem entirely the system may recom-mend a venue to a new user by looking at the contextualcircumstances and recommend the most preferred venuebased on these contextual circumstances In additioneven if a new venue is added to the system it can berecommended by looking at the content-related charac-teristics )e online evaluation was conducted with only20 people which is a relatively small sample for obtainingstatistically precise and coherent results )erefore moreusers should be reached for future studies Finally thisalgorithm can be embedded into a mobile application andthis mobile application can be marketed in mobile ap-plication stores )erefore its performance can be mea-sured via online experiments As users download and usethe application they will score the recommendation sothat online feedback can be taken and the algorithm can beimproved

10 Mobile Information Systems

Data Availability

)e Twitter and Foursquare data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

)e authors declare that they have no conflicts of interest

Acknowledgments

)is study was funded by Bogazici University ScientificResearch Fund (grant no 11463)

References

[1] A Bozanta and B Kutlu ldquoHybRecSys content-based con-textual hybrid venue recommender systemrdquo Journal of In-formation Science pp 1ndash15 2018

[2] M Sattari I H Toroslu P Karagoz P Symeonidis andY Manolopoulos ldquoExtended feature combination model forrecommendations in location-based mobile servicesrdquoKnowledge and Information Systems vol 44 no 3 pp 629ndash661 2015

[3] H Wang G Li and J Feng ldquoGroup-based personalizedlocation recommendation on social networksrdquo in WebTechnologies and Applications pp 68ndash80 Springer BerlinGermany 2014

[4] H Yin B Cui Y Sun Z Hu and L Chen ldquoLCARS a spatialitem recommender systemrdquo ACM Transactions on In-formation Systems (TOIS) vol 32 no 3 pp 1ndash37 2014

[5] M H Park J H Hong and S B Cho ldquoLocation-basedrecommendation system using bayesian userrsquos preferencemodel in mobile devicesrdquo in Ubiquitous Intelligence andComputing pp 1130ndash1139 Springer Berlin HeidelbergGermany 2007

[6] K Aihara H Koshiba and H Takeda ldquoBehavioral cost-basedrecommendation model for wanderers in townrdquo in Human-Computer Interaction Towards Mobile and IntelligentInteraction Environments pp 271ndash279 Springer BerlinHeidelberg Germany 2011

[7] Z Yu Y Feng H Xu and X Zhou ldquoRecommending travelpackages based on mobile crowdsourced datardquo IEEE Com-munications Magazine vol 52 no 8 pp 56ndash62 2014

[8] H Yin B Cui L Chen Z Hu and C Zhang ldquoModelinglocation-based user rating profiles for personalized recom-mendationrdquo ACM Transactions on Knowledge Discovery fromData vol 9 no 3 pp 1ndash41 2015

[9] D Zhou S M Rahimi and X Wang ldquoSimilarity-basedprobabilistic category-based location recommendation uti-lizing temporal and geographical influencerdquo InternationalJournal of Data Science and Analytics vol 1 no 2 pp 111ndash121 2016

[10] H Zhu E Chen H Xiong K Yu H Cao and J TianldquoMining mobile user preferences for personalized context-aware recommendationrdquo ACM Transactions on IntelligentSystems and Technology vol 5 no 4 pp 1ndash27 2015

[11] Q Fang C Xu M S Hossain and M G Stcaplrs ldquoA spatial-temporal context-aware personalized location recommen-dation systemrdquo ACM Transactions on Intelligent systems andTechnology vol 7 no 4 pp 1ndash30 2016

[12] Q Yuan G Cong K Zhao Z Ma and A Sun ldquoWho wherewhen and what a nonparametric bayesian approach to

context-aware recommendation and search for twitter usersrdquoACM Transactions on Information Systems vol 33 no 1pp 1ndash31 2015

[13] A Gupta and K Singh ldquoLocation based personalized res-taurant recommendation system for mobile environmentsrdquoin Proceedings of International Conference on Advances inComputing Communications and Informatics (ICACCI)pp 507ndash511 IEEE Mysore India August 2013

[14] J Bao Y Zheng and M F Mokbel ldquoLocation-based andpreference-aware recommendation using sparse geo-socialnetworking datardquo in Proceedings of 20th International Con-ference on Advances in Geographic Information Systemspp 199ndash208 Redondo Beach CA USA November 2012

[15] H Yin Y Sun B Cui Z Hu and L Chen ldquoLcars a location-content-aware recommender systemrdquo in Proceedings of 19thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 221ndash229 Chicago IL USAAugust 2013

[16] M H Kuo L C Chen and C W Liang ldquoBuilding andevaluating a location-based service recommendation systemwith a preference adjustment mechanismrdquo Expert Systemswith Applications vol 36 no 2 pp 3543ndash3554 2009

[17] K Shimada H Uehara and T Endo ldquoA comparative study ofpotential-of-interest days on a sightseeing spot recom-menderrdquo in Proceedings of IIAI 3rd International Conferenceon Advanced Applied Informatics (IIAIAAI) pp 555ndash560Kitakyushu Japan August 2014

[18] L Ahmedi K Rrmoku K Sylejmani and D Shabani ldquoAbimodal social network analysis to recommend points ofinterest to touristsrdquo Social Network Analysis and Miningvol 7 no 1 p 14 2017

[19] H Gao J Tang X Hu and H Liu ldquoContent-aware point ofinterest recommendation on location-based social networksrdquoin Proceedings of the AAAI pp 1721ndash1727 Austin TexasUSA January 2015

[20] Z Zheng Y Chen S Chen L Sun and D Chen ldquoLocation-aware POI recommendation for indoor space by exploitingWiFi logsrdquo Mobile Information Systems vol 2017 Article ID9601404 16 pages 2017

[21] W X Zhao N Zhou A Sun J R Wen J Han andE Y Chang ldquoA time-aware trajectory embedding model fornext-location recommendationrdquo Knowledge and InformationSystems vol 56 no 3 pp 559ndash579 2017

[22] R Srivastava S Hingmire G K Palshikar S Chaurasia andD A Csrs ldquoA context and sequence aware recommendationsystemrdquo in Proceedings of 8th Annual Meeting of the Forum onInformation Retrieval Evaluation pp 8ndash15 Kolkata IndiaDecember 2016

[23] V Subramaniyaswamy V Vijayakumar R Logesh andV Indragandhi ldquoIntelligent travel recommendation systemby mining attributes from community contributed photosrdquoProcedia Computer Science vol 50 pp 447ndash455 2015

[24] L Cao J Luo A C Gallagher X Jin J Han and T S HuangldquoA worldwide tourism recommendation system based ongeotaggedweb photosrdquo in Proceedings of ICASSP pp 2274ndash2277 Dallas Texas USA March 2010

[25] L Guo J Shao K L Tan and Y Y Wheretogo ldquoPersonalizedtravel recommendation for individuals and groupsrdquo in Pro-ceedings of IEEE 15th International Conference onMobile DataManagement vol 1 pp 49ndash58 Brisbane Australia July 2014

[26] I Memon L Chen A Majid M Lv I Hussain and G ChenldquoTravel recommendation using geo-tagged photos in socialmedia for touristrdquoWireless Personal Communications vol 80no 4 pp 1347ndash1362 2015

Mobile Information Systems 11

[27] Z Xu L Chen and G Chen ldquoTopic based context-awaretravel recommendation method exploiting geotaggedphotosrdquo Neurocomputing vol 155 pp 99ndash107 2015

[28] B Dhake S S Lomte R A Auti Y R Nagargoje and B PatilldquoLARS an efficient and scalable location-awarerdquo In-ternational Journal of Scientific Research And Educationvol 2 p 11 2014

[29] M Sarwat J J Levandoski A Eldawy and M F MokbelldquoLARSlowast an efficient and scalable location-aware recom-mender systemrdquo IEEE Transactions on Knowledge and DataEngineering vol 26 no 6 pp 1384ndash1399 2014

[30] P V Krishna S Misra D Joshi andM S Obaidat ldquoLearningautomata based sentiment analysis for recommender systemon cloudrdquo in Proceedings of International Conference onComputer Information and Telecommunication Systems(CITS) pp 1ndash5 IEEE Athens Greece May 2013

[31] M Mordacchini A Passarella M Conti et al ldquoCrowd-sourcing through cognitive opportunistic networksrdquo ACMTransactions on Autonomous and Adaptive Systems vol 10no 2 pp 1ndash29 2015

[32] Y Zheng L Zhang Z Ma X Xie and W Y Ma ldquoRec-ommending friends and locations based on individual loca-tion historyrdquo ACM Transactions on the Web vol 5 no 1pp 1ndash44 2011

[33] VW Zheng Y Zheng X Xie and Q Yang ldquoTowards mobileintelligence learning from GPS history data for collaborativerecommendationrdquo Artificial Intelligence vol 184ndash185pp 17ndash37 2012

[34] M Korakakis E Spyrou P Mylonas and S J PerantonisldquoExploiting social media information toward a context-awarerecommendation systemrdquo Social Network Analysis andMining vol 7 no 1 p 42 2017

[35] Q Li Y Zheng X Xie Y Chen W Liu and W Y MaldquoMining user similarity based on location historyrdquo in Pro-ceedings of 16th ACM SIGSPATIAL International Conferenceon Advances in Geographic Information Systems p 34 ACMNew York NY USA November 2008

[36] M Ye P Yin and W C Lee ldquoLocation recommendation forlocation-based social networksrdquo in Proceedings of 18th SIG-SPATIAL International Conference on Advances in GeographicInformation Systems pp 458ndash461 ACM San Jose CA USANovember 2010

[37] T Hasegawa and T Hayashi ldquoCollaborative filtering basedspot recommendation seamlessly available in home and awayareasrdquo in Proceedings of the IEEEACIS 12th InternationalConference on Computer and Information Science (ICIS)pp 547-548 IEEE Niigata Japan June 2013

[38] J J Levandoski M Sarwat A Eldawy and M F MokbelldquoLars a location-aware recommender systemrdquo in Proceedingsof IEEE 28th International Conference on Data Engineering(ICDE) pp 450ndash461 IEEE Arlington VA USA April 2012

[39] Y Takeuchi and M Sugimoto ldquoCityvoyager an outdoorrecommendation system based on user location historyrdquo inUbiquitous Intelligence and Computing pp 625ndash636Springer-Verlag Berlin Germany 2006

[40] S Knoch A Chapko A Emrich D Werth and P Loos ldquoAcontext-aware running route recommender learning fromuser histories using artificial neural networksrdquo in Proceedingsof 23rd International Workshop on Database and ExpertSystems Applications (DEXA) pp 106ndash110 IEEE ViennaAustria September 2012

[41] M Saraee S Khan and S Yamaner ldquoData mining approachto implement a recommendation system for electronic tourguidesrdquo in Proceedings of 2005 International Conference on

E-Business Enterprise Information Systems E-Governmentand Outsourcing EEE 2005 pp 215ndash218 CSREA PressLas Vegas NV USA June 2005

[42] R Katarya and O P Verma ldquoRestaurant recommendersystem based on psychographic and demographic factors inmobile environmentrdquo in Proceedings of International Con-ference on Green Computing and Internet of Dings (ICG-CIoT) pp 907ndash912 Uttar Pradesh India October 2015

[43] R Katarya and O P Verma ldquoRecent developments in af-fective recommender systemsrdquo Physica A Statistical Me-chanics and its Applications vol 461 pp 182ndash190 2016

[44] R Katarya ldquoA systematic review of group recommendersystems techniquesrdquo in Proceedings of International Confer-ence on Intelligent Sustainable Systems (ICISS) pp 425ndash428Tirupur India December 2017

[45] R Katarya O P Verma and I Jain ldquoUser behaviour analysisin context-aware recommender system using hybrid filteringapproachrdquo in Proceedings of 4th International Conference onComputer and Communication Technology (ICCCT)pp 222ndash227 Allahabad India September 2013

[46] R Katarya M Ranjan and O P Verma ldquoLocation basedrecommender system using enhanced random walk modelrdquoin Proceedings of Fourth International Conference on ParallelDistributed and Grid Computing (PDGC) pp 33ndash37Waknaghat India December 2016

[47] K Waga A Tabarcea and P Franti ldquoContext aware rec-ommendation of location-based datardquo in Proceedings of15th International Conference on System Deory Controland Computing (ICSTCC) pp 1ndash6 IEEE Sinaia RomaniaOctober 2011

[48] A Majid L Chen G Chen H T Mirza I Hussain andJ Woodward ldquoA context-aware personalized travel recom-mendation system based on geotagged social media dataminingrdquo International Journal of Geographical InformationScience vol 27 no 4 pp 662ndash684 2013

[49] Q Yuan G Cong Z Ma A Sun and N M)almann ldquoTime-aware point-of-interest recommendationrdquo in Proceedings of36th International ACM SIGIR Conference on Research andDevelopment in Information Retrieval pp 363ndash372 ACMDublin Ireland July 2013

[50] R Baral and T Li ldquoMaps a multi aspect personalized poirecommender systemrdquo in Proceedings of 10th ACM Confer-ence on Recommender Systems pp 281ndash284 ACM BostonMA USA September 2016

[51] P Hiesel M Braunhofer and W Worndl ldquoLearning thepopularity of items for mobile tourist guidesrdquo in Proceedingsof RecTour Cambridge MA USA September 2016

[52] L Yao Q Z Sheng X Wang W E Zhang and Y QinldquoCollaborative location recommendation by integratingmulti-dimensional contextual informationrdquo ACM Trans-actions on Internet Technology vol 18 no 3 pp 1ndash24 2018

[53] C Trattner A Oberegger L Eberhard D Parra andL Marinho ldquoUnderstanding the impact of weather for POIrecommendationsrdquo in Proceedings of RecTour CambridgeMA USA September 2016

[54] Y Zheng R Burke and B Mobasher ldquoDifferential contextrelaxation for context-aware travel recommendationrdquo inProceedings of International Conference on Electronic Com-merce and Web Technologies pp 88ndash99 Springer ViennaAustria September 2012

[55] M J Barranco J M Noguera J Castro and L Martinez ldquoAcontext-aware mobile recommender system based on locationand trajectoryrdquo in Management Intelligent Systems pp 153ndash162 Springer Berlin Heidelberg Germany 2012

12 Mobile Information Systems

[56] N S Savage M Baranski N E Chavez and T Hollerer ldquoIrsquomfeeling loco a location based context aware recommendationsystemrdquo in Advances in Location-Based Services pp 37ndash54Springer Berlin Heidelberg Germany 2012

[57] J L Herlocker J A Konstan L G Terveen and J T RiedlldquoEvaluating collaborative filtering recommender systemsrdquoACM Transactions on Information Systems vol 22 no 1pp 5ndash53 2004

[58] P Lops M De Gemmis and G Semeraro ldquoContent-basedrecommender systems state of the art and trendsrdquo in Rec-ommender Systems Handbook pp 73ndash105 Springer BerlinHeidelberg Germany 2011

[59] Z Lu H Wang N Mamoulis W Tu and D W CheungldquoPersonalized location recommendation by aggregatingmultiple recommenders in diversityrdquo GeoInformatica vol 21no 3 pp 459ndash484 2017

[60] R Burke ldquoHybrid recommender systems survey and ex-perimentsrdquo User Modeling and User-Adapted Interactionvol 12 no 4 pp 331ndash370 2002

[61] L Baltrunas and X Amatriain ldquoTowards time-dependentrecommendation based on implicit feedbackrdquo in Pro-ceedings of Workshop on Context-Aware Recommender Sys-tems (CARSrsquo09) New York NY USA October 2009

[62] X Wang Y L Zhao L Nie et al ldquoSemantic-based locationrecommendation with multimodal venue semanticsrdquo IEEETransactions on Multimedia vol 17 no 3 pp 409ndash4192015

[63] Q Yuan G Cong and A Sun ldquoGraph-based point-of-interest recommendation with geographical and temporalinfluencesrdquo in Proceedings of 23rd ACM International Con-ference on Information and Knowledge Managementpp 659ndash668 Shanghai China November 2014

[64] Y Mao P Yin W Lee and D Lee ldquoExploiting geographicalinfluence for collaborative point-of-interest recommenda-tionrdquo in Proceedings of 34th ACM SI- GIR InternationalConference on Research and Development in InformationRetrieval pp 325ndash334 Beijing China July 2011

[65] S K Gorakala and M Usuelli Building a RecommendationSystem with R Packt Publishing Ltd Birmingham UK 2015

[66] Y Zheng L Zhang X Xie andW Y Ma ldquoMining interestinglocations and travel sequences from GPS trajectoriesrdquo inProceedings of 18th International Conference on World WideWeb pp 791ndash800 Madrid Spain April 2009

[67] G Shani and A Gunawardana ldquoEvaluating recommendationsystemsrdquo in Recommender Systems Handbook pp 257ndash297Springer Berlin Heidelberg Germany 2011

[68] J A Konstan B N Miller D Maltz J L HerlockerL R Gordon and J Riedl ldquoGroupLens applying collabo-rative filtering to Usenet newsrdquo Communications of the ACMvol 40 no 3 pp 77ndash87 1997

[69] M Deshpande and G Karypis ldquoItem-based top-n recom-mendation algorithmsrdquo ACM Transactions on InformationSystems vol 22 no 1 pp 143ndash177 2004

[70] Y Koren R Bell and C Volinsky ldquoMatrix factorizationtechniques for recommender systemsrdquo Computer vol 42no 8 pp 30ndash37 2009

[71] Y Koren ldquoFactorization meets the neighborhood a multi-faceted collaborative filtering modelrdquo in Proceedings of 14thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 426ndash434 Las Vegas NV USAAugust 2008

Mobile Information Systems 13

Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems

Hindawiwwwhindawicom

Volume 2018

International Journal of

ReconfigurableComputing

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawiwwwhindawicom Volume 2018

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational Intelligence and Neuroscience

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018

Human-ComputerInteraction

Advances in

Hindawiwwwhindawicom Volume 2018

Scientic Programming

Submit your manuscripts atwwwhindawicom

Data Availability

)e Twitter and Foursquare data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

)e authors declare that they have no conflicts of interest

Acknowledgments

)is study was funded by Bogazici University ScientificResearch Fund (grant no 11463)

References

[1] A Bozanta and B Kutlu ldquoHybRecSys content-based con-textual hybrid venue recommender systemrdquo Journal of In-formation Science pp 1ndash15 2018

[2] M Sattari I H Toroslu P Karagoz P Symeonidis andY Manolopoulos ldquoExtended feature combination model forrecommendations in location-based mobile servicesrdquoKnowledge and Information Systems vol 44 no 3 pp 629ndash661 2015

[3] H Wang G Li and J Feng ldquoGroup-based personalizedlocation recommendation on social networksrdquo in WebTechnologies and Applications pp 68ndash80 Springer BerlinGermany 2014

[4] H Yin B Cui Y Sun Z Hu and L Chen ldquoLCARS a spatialitem recommender systemrdquo ACM Transactions on In-formation Systems (TOIS) vol 32 no 3 pp 1ndash37 2014

[5] M H Park J H Hong and S B Cho ldquoLocation-basedrecommendation system using bayesian userrsquos preferencemodel in mobile devicesrdquo in Ubiquitous Intelligence andComputing pp 1130ndash1139 Springer Berlin HeidelbergGermany 2007

[6] K Aihara H Koshiba and H Takeda ldquoBehavioral cost-basedrecommendation model for wanderers in townrdquo in Human-Computer Interaction Towards Mobile and IntelligentInteraction Environments pp 271ndash279 Springer BerlinHeidelberg Germany 2011

[7] Z Yu Y Feng H Xu and X Zhou ldquoRecommending travelpackages based on mobile crowdsourced datardquo IEEE Com-munications Magazine vol 52 no 8 pp 56ndash62 2014

[8] H Yin B Cui L Chen Z Hu and C Zhang ldquoModelinglocation-based user rating profiles for personalized recom-mendationrdquo ACM Transactions on Knowledge Discovery fromData vol 9 no 3 pp 1ndash41 2015

[9] D Zhou S M Rahimi and X Wang ldquoSimilarity-basedprobabilistic category-based location recommendation uti-lizing temporal and geographical influencerdquo InternationalJournal of Data Science and Analytics vol 1 no 2 pp 111ndash121 2016

[10] H Zhu E Chen H Xiong K Yu H Cao and J TianldquoMining mobile user preferences for personalized context-aware recommendationrdquo ACM Transactions on IntelligentSystems and Technology vol 5 no 4 pp 1ndash27 2015

[11] Q Fang C Xu M S Hossain and M G Stcaplrs ldquoA spatial-temporal context-aware personalized location recommen-dation systemrdquo ACM Transactions on Intelligent systems andTechnology vol 7 no 4 pp 1ndash30 2016

[12] Q Yuan G Cong K Zhao Z Ma and A Sun ldquoWho wherewhen and what a nonparametric bayesian approach to

context-aware recommendation and search for twitter usersrdquoACM Transactions on Information Systems vol 33 no 1pp 1ndash31 2015

[13] A Gupta and K Singh ldquoLocation based personalized res-taurant recommendation system for mobile environmentsrdquoin Proceedings of International Conference on Advances inComputing Communications and Informatics (ICACCI)pp 507ndash511 IEEE Mysore India August 2013

[14] J Bao Y Zheng and M F Mokbel ldquoLocation-based andpreference-aware recommendation using sparse geo-socialnetworking datardquo in Proceedings of 20th International Con-ference on Advances in Geographic Information Systemspp 199ndash208 Redondo Beach CA USA November 2012

[15] H Yin Y Sun B Cui Z Hu and L Chen ldquoLcars a location-content-aware recommender systemrdquo in Proceedings of 19thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 221ndash229 Chicago IL USAAugust 2013

[16] M H Kuo L C Chen and C W Liang ldquoBuilding andevaluating a location-based service recommendation systemwith a preference adjustment mechanismrdquo Expert Systemswith Applications vol 36 no 2 pp 3543ndash3554 2009

[17] K Shimada H Uehara and T Endo ldquoA comparative study ofpotential-of-interest days on a sightseeing spot recom-menderrdquo in Proceedings of IIAI 3rd International Conferenceon Advanced Applied Informatics (IIAIAAI) pp 555ndash560Kitakyushu Japan August 2014

[18] L Ahmedi K Rrmoku K Sylejmani and D Shabani ldquoAbimodal social network analysis to recommend points ofinterest to touristsrdquo Social Network Analysis and Miningvol 7 no 1 p 14 2017

[19] H Gao J Tang X Hu and H Liu ldquoContent-aware point ofinterest recommendation on location-based social networksrdquoin Proceedings of the AAAI pp 1721ndash1727 Austin TexasUSA January 2015

[20] Z Zheng Y Chen S Chen L Sun and D Chen ldquoLocation-aware POI recommendation for indoor space by exploitingWiFi logsrdquo Mobile Information Systems vol 2017 Article ID9601404 16 pages 2017

[21] W X Zhao N Zhou A Sun J R Wen J Han andE Y Chang ldquoA time-aware trajectory embedding model fornext-location recommendationrdquo Knowledge and InformationSystems vol 56 no 3 pp 559ndash579 2017

[22] R Srivastava S Hingmire G K Palshikar S Chaurasia andD A Csrs ldquoA context and sequence aware recommendationsystemrdquo in Proceedings of 8th Annual Meeting of the Forum onInformation Retrieval Evaluation pp 8ndash15 Kolkata IndiaDecember 2016

[23] V Subramaniyaswamy V Vijayakumar R Logesh andV Indragandhi ldquoIntelligent travel recommendation systemby mining attributes from community contributed photosrdquoProcedia Computer Science vol 50 pp 447ndash455 2015

[24] L Cao J Luo A C Gallagher X Jin J Han and T S HuangldquoA worldwide tourism recommendation system based ongeotaggedweb photosrdquo in Proceedings of ICASSP pp 2274ndash2277 Dallas Texas USA March 2010

[25] L Guo J Shao K L Tan and Y Y Wheretogo ldquoPersonalizedtravel recommendation for individuals and groupsrdquo in Pro-ceedings of IEEE 15th International Conference onMobile DataManagement vol 1 pp 49ndash58 Brisbane Australia July 2014

[26] I Memon L Chen A Majid M Lv I Hussain and G ChenldquoTravel recommendation using geo-tagged photos in socialmedia for touristrdquoWireless Personal Communications vol 80no 4 pp 1347ndash1362 2015

Mobile Information Systems 11

[27] Z Xu L Chen and G Chen ldquoTopic based context-awaretravel recommendation method exploiting geotaggedphotosrdquo Neurocomputing vol 155 pp 99ndash107 2015

[28] B Dhake S S Lomte R A Auti Y R Nagargoje and B PatilldquoLARS an efficient and scalable location-awarerdquo In-ternational Journal of Scientific Research And Educationvol 2 p 11 2014

[29] M Sarwat J J Levandoski A Eldawy and M F MokbelldquoLARSlowast an efficient and scalable location-aware recom-mender systemrdquo IEEE Transactions on Knowledge and DataEngineering vol 26 no 6 pp 1384ndash1399 2014

[30] P V Krishna S Misra D Joshi andM S Obaidat ldquoLearningautomata based sentiment analysis for recommender systemon cloudrdquo in Proceedings of International Conference onComputer Information and Telecommunication Systems(CITS) pp 1ndash5 IEEE Athens Greece May 2013

[31] M Mordacchini A Passarella M Conti et al ldquoCrowd-sourcing through cognitive opportunistic networksrdquo ACMTransactions on Autonomous and Adaptive Systems vol 10no 2 pp 1ndash29 2015

[32] Y Zheng L Zhang Z Ma X Xie and W Y Ma ldquoRec-ommending friends and locations based on individual loca-tion historyrdquo ACM Transactions on the Web vol 5 no 1pp 1ndash44 2011

[33] VW Zheng Y Zheng X Xie and Q Yang ldquoTowards mobileintelligence learning from GPS history data for collaborativerecommendationrdquo Artificial Intelligence vol 184ndash185pp 17ndash37 2012

[34] M Korakakis E Spyrou P Mylonas and S J PerantonisldquoExploiting social media information toward a context-awarerecommendation systemrdquo Social Network Analysis andMining vol 7 no 1 p 42 2017

[35] Q Li Y Zheng X Xie Y Chen W Liu and W Y MaldquoMining user similarity based on location historyrdquo in Pro-ceedings of 16th ACM SIGSPATIAL International Conferenceon Advances in Geographic Information Systems p 34 ACMNew York NY USA November 2008

[36] M Ye P Yin and W C Lee ldquoLocation recommendation forlocation-based social networksrdquo in Proceedings of 18th SIG-SPATIAL International Conference on Advances in GeographicInformation Systems pp 458ndash461 ACM San Jose CA USANovember 2010

[37] T Hasegawa and T Hayashi ldquoCollaborative filtering basedspot recommendation seamlessly available in home and awayareasrdquo in Proceedings of the IEEEACIS 12th InternationalConference on Computer and Information Science (ICIS)pp 547-548 IEEE Niigata Japan June 2013

[38] J J Levandoski M Sarwat A Eldawy and M F MokbelldquoLars a location-aware recommender systemrdquo in Proceedingsof IEEE 28th International Conference on Data Engineering(ICDE) pp 450ndash461 IEEE Arlington VA USA April 2012

[39] Y Takeuchi and M Sugimoto ldquoCityvoyager an outdoorrecommendation system based on user location historyrdquo inUbiquitous Intelligence and Computing pp 625ndash636Springer-Verlag Berlin Germany 2006

[40] S Knoch A Chapko A Emrich D Werth and P Loos ldquoAcontext-aware running route recommender learning fromuser histories using artificial neural networksrdquo in Proceedingsof 23rd International Workshop on Database and ExpertSystems Applications (DEXA) pp 106ndash110 IEEE ViennaAustria September 2012

[41] M Saraee S Khan and S Yamaner ldquoData mining approachto implement a recommendation system for electronic tourguidesrdquo in Proceedings of 2005 International Conference on

E-Business Enterprise Information Systems E-Governmentand Outsourcing EEE 2005 pp 215ndash218 CSREA PressLas Vegas NV USA June 2005

[42] R Katarya and O P Verma ldquoRestaurant recommendersystem based on psychographic and demographic factors inmobile environmentrdquo in Proceedings of International Con-ference on Green Computing and Internet of Dings (ICG-CIoT) pp 907ndash912 Uttar Pradesh India October 2015

[43] R Katarya and O P Verma ldquoRecent developments in af-fective recommender systemsrdquo Physica A Statistical Me-chanics and its Applications vol 461 pp 182ndash190 2016

[44] R Katarya ldquoA systematic review of group recommendersystems techniquesrdquo in Proceedings of International Confer-ence on Intelligent Sustainable Systems (ICISS) pp 425ndash428Tirupur India December 2017

[45] R Katarya O P Verma and I Jain ldquoUser behaviour analysisin context-aware recommender system using hybrid filteringapproachrdquo in Proceedings of 4th International Conference onComputer and Communication Technology (ICCCT)pp 222ndash227 Allahabad India September 2013

[46] R Katarya M Ranjan and O P Verma ldquoLocation basedrecommender system using enhanced random walk modelrdquoin Proceedings of Fourth International Conference on ParallelDistributed and Grid Computing (PDGC) pp 33ndash37Waknaghat India December 2016

[47] K Waga A Tabarcea and P Franti ldquoContext aware rec-ommendation of location-based datardquo in Proceedings of15th International Conference on System Deory Controland Computing (ICSTCC) pp 1ndash6 IEEE Sinaia RomaniaOctober 2011

[48] A Majid L Chen G Chen H T Mirza I Hussain andJ Woodward ldquoA context-aware personalized travel recom-mendation system based on geotagged social media dataminingrdquo International Journal of Geographical InformationScience vol 27 no 4 pp 662ndash684 2013

[49] Q Yuan G Cong Z Ma A Sun and N M)almann ldquoTime-aware point-of-interest recommendationrdquo in Proceedings of36th International ACM SIGIR Conference on Research andDevelopment in Information Retrieval pp 363ndash372 ACMDublin Ireland July 2013

[50] R Baral and T Li ldquoMaps a multi aspect personalized poirecommender systemrdquo in Proceedings of 10th ACM Confer-ence on Recommender Systems pp 281ndash284 ACM BostonMA USA September 2016

[51] P Hiesel M Braunhofer and W Worndl ldquoLearning thepopularity of items for mobile tourist guidesrdquo in Proceedingsof RecTour Cambridge MA USA September 2016

[52] L Yao Q Z Sheng X Wang W E Zhang and Y QinldquoCollaborative location recommendation by integratingmulti-dimensional contextual informationrdquo ACM Trans-actions on Internet Technology vol 18 no 3 pp 1ndash24 2018

[53] C Trattner A Oberegger L Eberhard D Parra andL Marinho ldquoUnderstanding the impact of weather for POIrecommendationsrdquo in Proceedings of RecTour CambridgeMA USA September 2016

[54] Y Zheng R Burke and B Mobasher ldquoDifferential contextrelaxation for context-aware travel recommendationrdquo inProceedings of International Conference on Electronic Com-merce and Web Technologies pp 88ndash99 Springer ViennaAustria September 2012

[55] M J Barranco J M Noguera J Castro and L Martinez ldquoAcontext-aware mobile recommender system based on locationand trajectoryrdquo in Management Intelligent Systems pp 153ndash162 Springer Berlin Heidelberg Germany 2012

12 Mobile Information Systems

[56] N S Savage M Baranski N E Chavez and T Hollerer ldquoIrsquomfeeling loco a location based context aware recommendationsystemrdquo in Advances in Location-Based Services pp 37ndash54Springer Berlin Heidelberg Germany 2012

[57] J L Herlocker J A Konstan L G Terveen and J T RiedlldquoEvaluating collaborative filtering recommender systemsrdquoACM Transactions on Information Systems vol 22 no 1pp 5ndash53 2004

[58] P Lops M De Gemmis and G Semeraro ldquoContent-basedrecommender systems state of the art and trendsrdquo in Rec-ommender Systems Handbook pp 73ndash105 Springer BerlinHeidelberg Germany 2011

[59] Z Lu H Wang N Mamoulis W Tu and D W CheungldquoPersonalized location recommendation by aggregatingmultiple recommenders in diversityrdquo GeoInformatica vol 21no 3 pp 459ndash484 2017

[60] R Burke ldquoHybrid recommender systems survey and ex-perimentsrdquo User Modeling and User-Adapted Interactionvol 12 no 4 pp 331ndash370 2002

[61] L Baltrunas and X Amatriain ldquoTowards time-dependentrecommendation based on implicit feedbackrdquo in Pro-ceedings of Workshop on Context-Aware Recommender Sys-tems (CARSrsquo09) New York NY USA October 2009

[62] X Wang Y L Zhao L Nie et al ldquoSemantic-based locationrecommendation with multimodal venue semanticsrdquo IEEETransactions on Multimedia vol 17 no 3 pp 409ndash4192015

[63] Q Yuan G Cong and A Sun ldquoGraph-based point-of-interest recommendation with geographical and temporalinfluencesrdquo in Proceedings of 23rd ACM International Con-ference on Information and Knowledge Managementpp 659ndash668 Shanghai China November 2014

[64] Y Mao P Yin W Lee and D Lee ldquoExploiting geographicalinfluence for collaborative point-of-interest recommenda-tionrdquo in Proceedings of 34th ACM SI- GIR InternationalConference on Research and Development in InformationRetrieval pp 325ndash334 Beijing China July 2011

[65] S K Gorakala and M Usuelli Building a RecommendationSystem with R Packt Publishing Ltd Birmingham UK 2015

[66] Y Zheng L Zhang X Xie andW Y Ma ldquoMining interestinglocations and travel sequences from GPS trajectoriesrdquo inProceedings of 18th International Conference on World WideWeb pp 791ndash800 Madrid Spain April 2009

[67] G Shani and A Gunawardana ldquoEvaluating recommendationsystemsrdquo in Recommender Systems Handbook pp 257ndash297Springer Berlin Heidelberg Germany 2011

[68] J A Konstan B N Miller D Maltz J L HerlockerL R Gordon and J Riedl ldquoGroupLens applying collabo-rative filtering to Usenet newsrdquo Communications of the ACMvol 40 no 3 pp 77ndash87 1997

[69] M Deshpande and G Karypis ldquoItem-based top-n recom-mendation algorithmsrdquo ACM Transactions on InformationSystems vol 22 no 1 pp 143ndash177 2004

[70] Y Koren R Bell and C Volinsky ldquoMatrix factorizationtechniques for recommender systemsrdquo Computer vol 42no 8 pp 30ndash37 2009

[71] Y Koren ldquoFactorization meets the neighborhood a multi-faceted collaborative filtering modelrdquo in Proceedings of 14thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 426ndash434 Las Vegas NV USAAugust 2008

Mobile Information Systems 13

Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems

Hindawiwwwhindawicom

Volume 2018

International Journal of

ReconfigurableComputing

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawiwwwhindawicom Volume 2018

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational Intelligence and Neuroscience

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018

Human-ComputerInteraction

Advances in

Hindawiwwwhindawicom Volume 2018

Scientic Programming

Submit your manuscripts atwwwhindawicom

[27] Z Xu L Chen and G Chen ldquoTopic based context-awaretravel recommendation method exploiting geotaggedphotosrdquo Neurocomputing vol 155 pp 99ndash107 2015

[28] B Dhake S S Lomte R A Auti Y R Nagargoje and B PatilldquoLARS an efficient and scalable location-awarerdquo In-ternational Journal of Scientific Research And Educationvol 2 p 11 2014

[29] M Sarwat J J Levandoski A Eldawy and M F MokbelldquoLARSlowast an efficient and scalable location-aware recom-mender systemrdquo IEEE Transactions on Knowledge and DataEngineering vol 26 no 6 pp 1384ndash1399 2014

[30] P V Krishna S Misra D Joshi andM S Obaidat ldquoLearningautomata based sentiment analysis for recommender systemon cloudrdquo in Proceedings of International Conference onComputer Information and Telecommunication Systems(CITS) pp 1ndash5 IEEE Athens Greece May 2013

[31] M Mordacchini A Passarella M Conti et al ldquoCrowd-sourcing through cognitive opportunistic networksrdquo ACMTransactions on Autonomous and Adaptive Systems vol 10no 2 pp 1ndash29 2015

[32] Y Zheng L Zhang Z Ma X Xie and W Y Ma ldquoRec-ommending friends and locations based on individual loca-tion historyrdquo ACM Transactions on the Web vol 5 no 1pp 1ndash44 2011

[33] VW Zheng Y Zheng X Xie and Q Yang ldquoTowards mobileintelligence learning from GPS history data for collaborativerecommendationrdquo Artificial Intelligence vol 184ndash185pp 17ndash37 2012

[34] M Korakakis E Spyrou P Mylonas and S J PerantonisldquoExploiting social media information toward a context-awarerecommendation systemrdquo Social Network Analysis andMining vol 7 no 1 p 42 2017

[35] Q Li Y Zheng X Xie Y Chen W Liu and W Y MaldquoMining user similarity based on location historyrdquo in Pro-ceedings of 16th ACM SIGSPATIAL International Conferenceon Advances in Geographic Information Systems p 34 ACMNew York NY USA November 2008

[36] M Ye P Yin and W C Lee ldquoLocation recommendation forlocation-based social networksrdquo in Proceedings of 18th SIG-SPATIAL International Conference on Advances in GeographicInformation Systems pp 458ndash461 ACM San Jose CA USANovember 2010

[37] T Hasegawa and T Hayashi ldquoCollaborative filtering basedspot recommendation seamlessly available in home and awayareasrdquo in Proceedings of the IEEEACIS 12th InternationalConference on Computer and Information Science (ICIS)pp 547-548 IEEE Niigata Japan June 2013

[38] J J Levandoski M Sarwat A Eldawy and M F MokbelldquoLars a location-aware recommender systemrdquo in Proceedingsof IEEE 28th International Conference on Data Engineering(ICDE) pp 450ndash461 IEEE Arlington VA USA April 2012

[39] Y Takeuchi and M Sugimoto ldquoCityvoyager an outdoorrecommendation system based on user location historyrdquo inUbiquitous Intelligence and Computing pp 625ndash636Springer-Verlag Berlin Germany 2006

[40] S Knoch A Chapko A Emrich D Werth and P Loos ldquoAcontext-aware running route recommender learning fromuser histories using artificial neural networksrdquo in Proceedingsof 23rd International Workshop on Database and ExpertSystems Applications (DEXA) pp 106ndash110 IEEE ViennaAustria September 2012

[41] M Saraee S Khan and S Yamaner ldquoData mining approachto implement a recommendation system for electronic tourguidesrdquo in Proceedings of 2005 International Conference on

E-Business Enterprise Information Systems E-Governmentand Outsourcing EEE 2005 pp 215ndash218 CSREA PressLas Vegas NV USA June 2005

[42] R Katarya and O P Verma ldquoRestaurant recommendersystem based on psychographic and demographic factors inmobile environmentrdquo in Proceedings of International Con-ference on Green Computing and Internet of Dings (ICG-CIoT) pp 907ndash912 Uttar Pradesh India October 2015

[43] R Katarya and O P Verma ldquoRecent developments in af-fective recommender systemsrdquo Physica A Statistical Me-chanics and its Applications vol 461 pp 182ndash190 2016

[44] R Katarya ldquoA systematic review of group recommendersystems techniquesrdquo in Proceedings of International Confer-ence on Intelligent Sustainable Systems (ICISS) pp 425ndash428Tirupur India December 2017

[45] R Katarya O P Verma and I Jain ldquoUser behaviour analysisin context-aware recommender system using hybrid filteringapproachrdquo in Proceedings of 4th International Conference onComputer and Communication Technology (ICCCT)pp 222ndash227 Allahabad India September 2013

[46] R Katarya M Ranjan and O P Verma ldquoLocation basedrecommender system using enhanced random walk modelrdquoin Proceedings of Fourth International Conference on ParallelDistributed and Grid Computing (PDGC) pp 33ndash37Waknaghat India December 2016

[47] K Waga A Tabarcea and P Franti ldquoContext aware rec-ommendation of location-based datardquo in Proceedings of15th International Conference on System Deory Controland Computing (ICSTCC) pp 1ndash6 IEEE Sinaia RomaniaOctober 2011

[48] A Majid L Chen G Chen H T Mirza I Hussain andJ Woodward ldquoA context-aware personalized travel recom-mendation system based on geotagged social media dataminingrdquo International Journal of Geographical InformationScience vol 27 no 4 pp 662ndash684 2013

[49] Q Yuan G Cong Z Ma A Sun and N M)almann ldquoTime-aware point-of-interest recommendationrdquo in Proceedings of36th International ACM SIGIR Conference on Research andDevelopment in Information Retrieval pp 363ndash372 ACMDublin Ireland July 2013

[50] R Baral and T Li ldquoMaps a multi aspect personalized poirecommender systemrdquo in Proceedings of 10th ACM Confer-ence on Recommender Systems pp 281ndash284 ACM BostonMA USA September 2016

[51] P Hiesel M Braunhofer and W Worndl ldquoLearning thepopularity of items for mobile tourist guidesrdquo in Proceedingsof RecTour Cambridge MA USA September 2016

[52] L Yao Q Z Sheng X Wang W E Zhang and Y QinldquoCollaborative location recommendation by integratingmulti-dimensional contextual informationrdquo ACM Trans-actions on Internet Technology vol 18 no 3 pp 1ndash24 2018

[53] C Trattner A Oberegger L Eberhard D Parra andL Marinho ldquoUnderstanding the impact of weather for POIrecommendationsrdquo in Proceedings of RecTour CambridgeMA USA September 2016

[54] Y Zheng R Burke and B Mobasher ldquoDifferential contextrelaxation for context-aware travel recommendationrdquo inProceedings of International Conference on Electronic Com-merce and Web Technologies pp 88ndash99 Springer ViennaAustria September 2012

[55] M J Barranco J M Noguera J Castro and L Martinez ldquoAcontext-aware mobile recommender system based on locationand trajectoryrdquo in Management Intelligent Systems pp 153ndash162 Springer Berlin Heidelberg Germany 2012

12 Mobile Information Systems

[56] N S Savage M Baranski N E Chavez and T Hollerer ldquoIrsquomfeeling loco a location based context aware recommendationsystemrdquo in Advances in Location-Based Services pp 37ndash54Springer Berlin Heidelberg Germany 2012

[57] J L Herlocker J A Konstan L G Terveen and J T RiedlldquoEvaluating collaborative filtering recommender systemsrdquoACM Transactions on Information Systems vol 22 no 1pp 5ndash53 2004

[58] P Lops M De Gemmis and G Semeraro ldquoContent-basedrecommender systems state of the art and trendsrdquo in Rec-ommender Systems Handbook pp 73ndash105 Springer BerlinHeidelberg Germany 2011

[59] Z Lu H Wang N Mamoulis W Tu and D W CheungldquoPersonalized location recommendation by aggregatingmultiple recommenders in diversityrdquo GeoInformatica vol 21no 3 pp 459ndash484 2017

[60] R Burke ldquoHybrid recommender systems survey and ex-perimentsrdquo User Modeling and User-Adapted Interactionvol 12 no 4 pp 331ndash370 2002

[61] L Baltrunas and X Amatriain ldquoTowards time-dependentrecommendation based on implicit feedbackrdquo in Pro-ceedings of Workshop on Context-Aware Recommender Sys-tems (CARSrsquo09) New York NY USA October 2009

[62] X Wang Y L Zhao L Nie et al ldquoSemantic-based locationrecommendation with multimodal venue semanticsrdquo IEEETransactions on Multimedia vol 17 no 3 pp 409ndash4192015

[63] Q Yuan G Cong and A Sun ldquoGraph-based point-of-interest recommendation with geographical and temporalinfluencesrdquo in Proceedings of 23rd ACM International Con-ference on Information and Knowledge Managementpp 659ndash668 Shanghai China November 2014

[64] Y Mao P Yin W Lee and D Lee ldquoExploiting geographicalinfluence for collaborative point-of-interest recommenda-tionrdquo in Proceedings of 34th ACM SI- GIR InternationalConference on Research and Development in InformationRetrieval pp 325ndash334 Beijing China July 2011

[65] S K Gorakala and M Usuelli Building a RecommendationSystem with R Packt Publishing Ltd Birmingham UK 2015

[66] Y Zheng L Zhang X Xie andW Y Ma ldquoMining interestinglocations and travel sequences from GPS trajectoriesrdquo inProceedings of 18th International Conference on World WideWeb pp 791ndash800 Madrid Spain April 2009

[67] G Shani and A Gunawardana ldquoEvaluating recommendationsystemsrdquo in Recommender Systems Handbook pp 257ndash297Springer Berlin Heidelberg Germany 2011

[68] J A Konstan B N Miller D Maltz J L HerlockerL R Gordon and J Riedl ldquoGroupLens applying collabo-rative filtering to Usenet newsrdquo Communications of the ACMvol 40 no 3 pp 77ndash87 1997

[69] M Deshpande and G Karypis ldquoItem-based top-n recom-mendation algorithmsrdquo ACM Transactions on InformationSystems vol 22 no 1 pp 143ndash177 2004

[70] Y Koren R Bell and C Volinsky ldquoMatrix factorizationtechniques for recommender systemsrdquo Computer vol 42no 8 pp 30ndash37 2009

[71] Y Koren ldquoFactorization meets the neighborhood a multi-faceted collaborative filtering modelrdquo in Proceedings of 14thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 426ndash434 Las Vegas NV USAAugust 2008

Mobile Information Systems 13

Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems

Hindawiwwwhindawicom

Volume 2018

International Journal of

ReconfigurableComputing

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawiwwwhindawicom Volume 2018

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational Intelligence and Neuroscience

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018

Human-ComputerInteraction

Advances in

Hindawiwwwhindawicom Volume 2018

Scientic Programming

Submit your manuscripts atwwwhindawicom

[56] N S Savage M Baranski N E Chavez and T Hollerer ldquoIrsquomfeeling loco a location based context aware recommendationsystemrdquo in Advances in Location-Based Services pp 37ndash54Springer Berlin Heidelberg Germany 2012

[57] J L Herlocker J A Konstan L G Terveen and J T RiedlldquoEvaluating collaborative filtering recommender systemsrdquoACM Transactions on Information Systems vol 22 no 1pp 5ndash53 2004

[58] P Lops M De Gemmis and G Semeraro ldquoContent-basedrecommender systems state of the art and trendsrdquo in Rec-ommender Systems Handbook pp 73ndash105 Springer BerlinHeidelberg Germany 2011

[59] Z Lu H Wang N Mamoulis W Tu and D W CheungldquoPersonalized location recommendation by aggregatingmultiple recommenders in diversityrdquo GeoInformatica vol 21no 3 pp 459ndash484 2017

[60] R Burke ldquoHybrid recommender systems survey and ex-perimentsrdquo User Modeling and User-Adapted Interactionvol 12 no 4 pp 331ndash370 2002

[61] L Baltrunas and X Amatriain ldquoTowards time-dependentrecommendation based on implicit feedbackrdquo in Pro-ceedings of Workshop on Context-Aware Recommender Sys-tems (CARSrsquo09) New York NY USA October 2009

[62] X Wang Y L Zhao L Nie et al ldquoSemantic-based locationrecommendation with multimodal venue semanticsrdquo IEEETransactions on Multimedia vol 17 no 3 pp 409ndash4192015

[63] Q Yuan G Cong and A Sun ldquoGraph-based point-of-interest recommendation with geographical and temporalinfluencesrdquo in Proceedings of 23rd ACM International Con-ference on Information and Knowledge Managementpp 659ndash668 Shanghai China November 2014

[64] Y Mao P Yin W Lee and D Lee ldquoExploiting geographicalinfluence for collaborative point-of-interest recommenda-tionrdquo in Proceedings of 34th ACM SI- GIR InternationalConference on Research and Development in InformationRetrieval pp 325ndash334 Beijing China July 2011

[65] S K Gorakala and M Usuelli Building a RecommendationSystem with R Packt Publishing Ltd Birmingham UK 2015

[66] Y Zheng L Zhang X Xie andW Y Ma ldquoMining interestinglocations and travel sequences from GPS trajectoriesrdquo inProceedings of 18th International Conference on World WideWeb pp 791ndash800 Madrid Spain April 2009

[67] G Shani and A Gunawardana ldquoEvaluating recommendationsystemsrdquo in Recommender Systems Handbook pp 257ndash297Springer Berlin Heidelberg Germany 2011

[68] J A Konstan B N Miller D Maltz J L HerlockerL R Gordon and J Riedl ldquoGroupLens applying collabo-rative filtering to Usenet newsrdquo Communications of the ACMvol 40 no 3 pp 77ndash87 1997

[69] M Deshpande and G Karypis ldquoItem-based top-n recom-mendation algorithmsrdquo ACM Transactions on InformationSystems vol 22 no 1 pp 143ndash177 2004

[70] Y Koren R Bell and C Volinsky ldquoMatrix factorizationtechniques for recommender systemsrdquo Computer vol 42no 8 pp 30ndash37 2009

[71] Y Koren ldquoFactorization meets the neighborhood a multi-faceted collaborative filtering modelrdquo in Proceedings of 14thACM SIGKDD International Conference on Knowledge Dis-covery and Data Mining pp 426ndash434 Las Vegas NV USAAugust 2008

Mobile Information Systems 13

Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems

Hindawiwwwhindawicom

Volume 2018

International Journal of

ReconfigurableComputing

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawiwwwhindawicom Volume 2018

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational Intelligence and Neuroscience

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018

Human-ComputerInteraction

Advances in

Hindawiwwwhindawicom Volume 2018

Scientic Programming

Submit your manuscripts atwwwhindawicom

Computer Games Technology

International Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Journal ofEngineeringVolume 2018

Advances in

FuzzySystems

Hindawiwwwhindawicom

Volume 2018

International Journal of

ReconfigurableComputing

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

thinspArtificial Intelligence

Hindawiwwwhindawicom Volumethinsp2018

Hindawiwwwhindawicom Volume 2018

Civil EngineeringAdvances in

Hindawiwwwhindawicom Volume 2018

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawiwwwhindawicom Volume 2018

Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational Intelligence and Neuroscience

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018

Human-ComputerInteraction

Advances in

Hindawiwwwhindawicom Volume 2018

Scientic Programming

Submit your manuscripts atwwwhindawicom