analysis of rural tourism demand characteristics and

Research ArticleAnalysis of Rural Tourism Demand Characteristics andExperience Differences Based on Association Rule Mining

Qun Jiang

Wuhan Polytechnic, Wuhan, Hubei 430074, China

Correspondence should be addressed to Qun Jiang; [email protected]

Received 25 September 2021; Revised 10 October 2021; Accepted 12 October 2021; Published 13 November 2021

Academic Editor: Xin Ning

Copyright © 2021 Qun Jiang. This is an open access article distributed under the Creative Commons Attribution License, whichpermits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Whether the characteristics of rural tourism changes or not provides the scale and basis for judging whether the rural tourismlandscape has changed, but it cannot provide a judgment on the impact of rural tourism landscape changes. The impact isrelative to the rural tourism landscape goal. The determination of rural tourism landscape objectives provides a baselinefor judging the direction and impact of rural tourism characteristics and provides a prerequisite for rural tourismlandscape actions. The determination of the quality target of the rural tourism landscape is mainly determined by theinternal process and external demand of the rural tourism landscape. Through in-depth research on the frequent patterngrowth algorithm FP-Growth, the algorithm can find frequent item sets by not generating candidate item sets. The core ofthe algorithm is the frequent pattern tree FP-tree, which can efficiently compress the transaction database. Based on theadvantages of FP-tree, this paper improves a FP_Apriori algorithm based on frequent pattern trees. This algorithm projectsthe entire transaction database onto the FP-tree, avoiding a lot of I/O overhead. At the same time, I propose a moredirectional and targeted search strategy for FP-tree, which reduces the running time of the algorithm and uses theprinciple of the Mapping_Apriori algorithm to prethin the frequent item sets. This article uses the text analysis method ofnetwork data to excavate the characteristics and internal structure of rural tourism demand. The rural tourism market hasa wide range of needs and multiple levels, and traditional research methods such as questionnaires have limited samplesize and sample structure. With the help of network data, text mining, and other statistical analysis methods, in-depthempirical research on the characteristics and spatial structure of rural tourism in a certain region can cover more researchgroups. The research confirms that the results of using text analysis and questionnaire analysis on the perception ofdestination image are relatively consistent. Therefore, the network text analysis method is an effective tool to study thedemand structure of the rural tourism market.

1. Introduction

Data mining is to dig out hidden, unknown relationships,patterns, or trends that have potential value to decisionmakers from large-scale heterogeneous or multisource dataand use these knowledge and laws to establish auxiliarydecision-making or prediction models [1]. It is the processof using various analysis tools to discover models or poten-tial relationships among massive amounts of data. Becausedata mining can find useful information in data and makeenterprises profitable, this makes the application of datamining technology more and more common. The com-monly used methods for data analysis using data mining

mainly include classification, regression analysis, clustering,association rules, time series patterns, and deviation analysis,which mine data from different perspectives [2]. Amongthem, the association rule technology is a classic data miningmethod, which can discover the direct potential interrela-tionships of data, and this relationship does not need to bedirectly expressed in the database [3]. The task of associationanalysis is to discover the degree of association or associa-tion rules between things [4].

In recent years, some places in China have recognizedthe development law of local rural tourism [5]. After fullyunderstanding the spatial laws of their rural tourism, citieshave replanned, adjusted, and optimized the spatial pattern

HindawiWireless Communications and Mobile ComputingVolume 2021, Article ID 8742950, 12 pageshttps://doi.org/10.1155/2021/8742950

https://orcid.org/0000-0003-0181-8659

https://creativecommons.org/licenses/by/4.0/

https://doi.org/10.1155/2021/8742950

of rural tourism and proposed strategies or countermeasuresfor the development of rural tourism space according tolocal conditions [6]. Studying the spatial change of rural tour-ism is helpful to discover the process of agglomeration and dif-fusion of tourism economic activities in the region; find the corenode, development axis, or development zone; and establish aregional tourism space that can integrate social, cultural, eco-nomic, and environmental resources. By constructing ruraltourism centers of different levels and levels based on the cen-trality of the rural tourism spatial network and formulating cor-responding development strategies, it is possible to coordinatethe rational development of regional rural tourism and integratetourism, leisure, specialty food production and consumption,and e-commerce [7]. Studying the spatial changes of rural tour-ism is helpful to develop and construct new tourism infrastruc-ture and to selectively retain existing facilities to realize scientificrural tourism spatial planning [8].Whether it is the transforma-tion and construction of rural tourism consulting centers, tour-ism distribution centers, or rural tourism landscape tourismvillages with rural characteristics, the spatial layout and plan-ning of rural tourism are always the constraints of the construc-tion of these centers and characteristic villages and towns, andthey are the effective use of different types of centers [9].

Aiming at the improvement of the Apriori algorithm, thispaper mainly proposes the FP_Apriori algorithm based onthe frequent pattern tree. The improvement ideas of the algo-rithm are mainly reflected in three aspects. First, in view ofthe shortcomings of traditional algorithms that scan the trans-action database multiple times, the frequent pattern tree is usedto store transactions, so that only one traversal scan of thetransaction database is required. Second, I optimize the com-plexity coefficient of the frequency calculation of the candidateitem set to improve the running speed and mining efficiency ofthe algorithm. Third, I take advantage of the nature of frequentitem sets to slim down Lk−1 and remove some useless candidateitem sets in advance. In addition, this paper makes a verydetailed research and analysis on the improvement ideas ofthe FP_Apriori algorithm, analyzes the optimization strategyof FP-tree and candidate set frequency calculation, and ana-lyzes the core steps of the algorithm. With the help of morethan 20,000 online online review data on 5 online platformsfrom 50 major rural tourist attractions in a certain region, thispaper explores the specific research content of the seasonalcycle, demand attributes, supply characteristics, and supply-demand gap of rural tourism market in a certain region. Theempirical research reveals that the demand enthusiasm of therural tourism market in a certain region has obvious dual-peak characteristics, insufficient demand upgrades, and thehomogenization and simplification of tourism supply content.The overall satisfaction of the market is relatively good, thedemand loss trend is obvious.

2. Related Work

A key problem in association rule mining is to find all fre-quent item sets that meet the minimum support given bythe user from the transaction set [10]. This step concentratesall the calculations. The original algorithm to solve the asso-ciation rule problem is the AIS algorithm. In order to

improve the AIS algorithm, related scholars have proposedthe OCD algorithm [11]. The OCD algorithm uses the com-bined information of the previous search to reduce the num-ber of candidate item sets produced this time. Later,researchers proposed the most famous Apriori algorithm inassociation rule mining and its variants AprioriTid andAprioriHybrid algorithms to discover frequent item sets[12]. Since then, many scholars have proposed algorithmsfor discovering frequent item sets of association rules, butmost of the algorithms are variants or improvements of theApriori algorithm [13, 14]. Since the Apriori algorithm is amultipass search algorithm, for a large data collection, theexternal memory must be read once for each search, and theI/O overhead is very high [15]. Therefore, most of theimproved algorithms are making a fuss on how to reduce thenumber of searches. In fact, what really affects the efficiencyof the classic association rule frequent item set discovery algo-rithm based on the Apriori algorithm is the calculation of theitem set and its support. If the number of different itemsincluded in the transaction data set is n, the frequent itemset discovery algorithm based on Apriori will calculate 2n itemsets. When n is relatively large, a combinatorial explosion willoccur. In fact, this will be an NP-hard problem.

The sampling algorithm randomly samples a part of thesample from the original data set and uses the sample tomine association rules to reduce the number of searches ofthe algorithm. However, because the data set often hasuneven data distribution, random sampling cannot guaran-tee that it can be drawn at all. Although the Partition algo-rithm reduces the burden of I/O by separately mining thedata collection partitions and finally summarizes it, it actu-ally increases the burden of the CPU. The DIC algorithmuses a dynamic calculation strategy to reduce the numberof searches to improve the efficiency of the algorithm, butthere is no fundamental difference in thinking from theApriori algorithm, and it is also a multi-pass search algo-rithm. These algorithms generate candidate item sets whenreading in transaction data, and generate many unnecessarycandidate item sets, which are computationally intensive.Especially for massive data sets, the above algorithms canonly have a certain mining efficiency under high minimumsupport and minimum credibility or after other constraintsare added. Otherwise, a combination explosion of frequentitem sets will occur. The inefficiency even exceeds the stor-age and computing power of the machine. Because any algo-rithm must calculate the item set and its support, what reallyaffects the efficiency of the algorithm is the calculation of theitem set and its support. Each calculation not only takes a lotof CPU time but also involves I/O requests. Therefore, onlyby essentially reducing the generation of unnecessary itemsets and reducing the calculation time for the support ofitem sets can the number of frequent item set generationand data mining time be greatly reduced.

The academic circle mainly studies the spatial distributioncharacteristics of rural tourism from the aspects of integrateddestination management, location advantages, and resourceenvironment [16]. For the integrated management of touristdestinations and the optimal allocation of resources, relatedscholars have studied the central structure of Korean rural

2 Wireless Communications and Mobile Computing

space and its corresponding development strategies based onthe resources of rural convenience facilities [17]. Advantageouslocations or traffic arteries have led to changes in the spatialpattern of rural tourism. Researchers reviewed tourism devel-opment studies and found that rural areas on the fringe of citiesstrongly attract day-trip tourists, and the surrounding areasattract low-level tourists [18]. Relevant scholars have studiedthe relationship between high-speed rail and tourist attractionbetween Perpignan, France, and Barcelona, Spain, and foundthat high-speed rail reduces transportation costs, improvesthe accessibility of destinations, and enhances the spatial com-petitiveness between destinations [19]. It promotes the concen-tration of tourism activities in Barcelona, but it is detrimental toPerpignan. In addition, some scholars have studied the spatialrelationship between the comfortable resources of culturalfacilities in Korea, the spatial relationship between the hot tubin the Appalachian Mountains, Ohio, and the suitability of for-est tourism land and found out how to influence the superiortourism resources or superior environment in the region [20].The tourism policy and development planning strategy of thedevelopment department will affect the spatial layout of ruraltourism, and so on. Research on the spatial distribution pat-terns of foreign scholars does not directly start from the spatialdistribution or structure of rural tourism resources or scenicspots but indirect research from the hierarchical system, attrac-tive conditions, development conditions, and population distri-bution of rural tourism resources or scenic spots [21].

Changes in rural tourism have an impact on the develop-ment positioning, industrial structure, and employment ofthe entire tourism industry. Research believes that tourism isa policy choice of local governments and a more general partialresponse of rural areas to the ever-changing continental andglobal economy [22]. As a new economic activity that replacesagriculture, tourism has changed the agricultural structure,restructured the countryside, and changed government deci-sions and long-term family decisions, directly affecting theorganizational structure of rural communities, environmentalquality, and inter-city relations. In fact, the integration effectof rural space is very obvious. Rural space is no longer purelyan area related to the production of agricultural products butis regarded as a place to stimulate new social and economicactivities. It usually integrates tourism, leisure, and specialtyfood. The large-scale development of tourism in the peripheryof remote villages can promote the transfer of remote rurallabor force and create tourism economic links with poor fami-lies. However, it is worth noting that tourism is a low-risk, low-return livelihood strategy. It can be seen that the developmentof rural tourism is not only a policy choice for the governmentin the face of global restructuring and local response but also away of economic activity and employment problems.

3. The Characteristics of Rural Tourism as aSpatial Analysis to Manage the Change ofRural Tourism Landscape Experience

3.1. The Overall Framework of the Classification of RuralTourism Characteristics. In order to continuously promotethe understanding of rural tourism landscape and the man-

agement of rural tourism landscape, it is very important toestablish a frame of reference that can communicate withrural tourism landscape. However, the classification processof “rural tourism landscape” is very complicated, becausethis object contains multiple dimensions, as well as humanperception and physical reality. The rural tourism landscapeis composed of components that appear in the “view field,”including landforms, water bodies, vegetation, and infra-structure. These are often referred to as rural tourism land-scape layers. Although it is very important to decomposethe rural tourism landscape into different layers, due to thenature of the integrity and complexity of the rural tourismlandscape, the interaction of these layers forms the ruraltourism landscape. The rural tourism landscape as a wholeoperates under a nested scale and needs to be linked to theadministrative management level. The DPSIR frameworkof rural tourism landscape changes is shown in Figure 1.

The classification system of rural tourism characteristicsis a parallel structural relationship, which is a flexible frame-work that can be continuously expanded. The dimensionaldivision is the choice of perspective for observing, under-standing, and regulating the object of the rural tourism land-scape. The selection of dimensionality is closely related tothe goal, level, and scale. The more dimensions are divided,the more thorough and comprehensive the understandingof rural tourism landscape. Therefore, the classification sys-tem of rural tourism characteristics can be applied to differ-ent spatial structure levels and administrative managementsystems. Whether it is a transnational scale, a national orregional scale, or a local scale, a hierarchical nesting relation-ship is formed between different scales.

From the perspective of linguistics, corresponding to theplanning process, the classification of rural tourism character-istics as a specific planning language also has three purposes orthree modes. The first is the instruction model—the intuitivephenomenon that describes the characteristics of rural tour-ism; the second is the evaluation model that expresses subjec-tive value judgments—the evaluation of the value and qualityof the characteristics of rural tourism; the third is the pre-scribed mode that requires others to act—regulate the futurevillages’ goals and methods of tourism characteristics. Thereis a progressive relationship among the three models, but theyshould be based on common problems or around commongoals, such as the alienation and decline of rural and ruraltourism landscapes. First, the characteristics of rural and ruraltourism need to be defined, analyzed, and described. Andthen, you make corresponding evaluations on the qualityand value of rural tourism characteristics from the perspectiveof human use and environmental impact. Finally, based on thegoals of sustainable development, the quality objectives ofrural tourism landscapes are stipulated, so as to form corre-sponding guiding strategies (protection, management, orplanning) to protect, strengthen, or restore the characteristicsof rural tourism. The basic purpose of the classification ofrural tourism characteristics is to serve the planning and man-agement of the rural tourism landscape.

3.2. Evaluation and Decision-Making of Rural TourismCharacteristics. The evaluation stage of rural tourism

3Wireless Communications and Mobile Computing

characteristics is a relatively subjective process, which isbased on human goals and values to evaluate the quality orvalue of rural tourism landscapes, including evaluationmethods and evaluation results, which are used to assistdecision-making, such as whether the development projectis allowed. The evaluation of rural tourism landscape focuseson the subject of evaluation. This is obviously different fromthe purpose of classification and description of rural tourismcharacteristics. The core of the evaluation model is that there

are very different values in different development stages andsocial backgrounds.

The evaluation of the rural tourism landscape not onlyincludes the characteristics of rural tourism but also involvesmany aspects, such as the environmental capacity of therural tourism landscape, the value of the rural tourism land-scape, the sensitivity of the rural tourism landscape, and thequality of the rural tourism landscape. In specific research,the content and methods of evaluation vary with the specific

Driving force

Social economy

Policy means

D emocracy

Climatechange

Humanbehaviour

Travel

Agriculture

IndustryService

Energy

Pressure

Land use change

Use of pesticides

Land expansion

Irrigation system

Intensification

State

Land cover change

footprint

Structure

Diversity

Correlation

Openness

Closure

Quality

Influence

Land use change impact

Loss of structural elements

Repair

Soil Erosion

Land abandonment

Flood event

Respond

Agricultural environment

plan

Action plan

Landscape convention

Mitigation and compensation

Policy/planning

Technology/countermeasure

Regeneration

.

Figure 1: The DPSIR framework of rural tourism landscape changes.


problems and the goals that must be met and the objects ofuse. The purpose of the evaluation is to ensure that thechanges in land use and the planning and design of develop-ment projects achieve a harmonious relationship with thesurrounding environment, or to further strengthen or shapenew rural tourism landscapes. The evaluation methodshould be as objective as possible, and establish clear logicand ensure its transparency. Inevitable subjective evaluation(such as the value of rural tourism landscape) requires theparticipation of professionals, and should be combined withthe historical characteristics of the rural tourism landscape.This process should ensure the participation of stakeholdersas much as possible, and the extent and nature of their par-ticipation should be clarified. The process of rural tourismfeature evaluation is shown in Figure 2.

Based on the analysis and evaluation of rural tourismcharacteristics, the evaluation conclusions can be drawn toassist decision-making, including the determination of ruraltourism landscape quality objectives for each rural tourismlandscape type or region and corresponding rural tourismlandscape actions. The type of evaluation conclusion is alsoclosely related to the specific user population.

In the decision-making stage of the evaluation of ruraltourism characteristics, based on the results of investigation,analysis, and evaluation, the subjective judgment of theresearcher, the application method, and the opinions ofstakeholders are added, that is, “who” will implement deter-mines the development of the decision-making stage. Thisprocess will inevitably have various subjectivities, so a clearprocess and causality are very important. In the process ofevaluation formation and the use of evaluation results, it isnecessary to understand which elements are relatively objec-tive and noncontroversial and which elements are prone toproduce different opinions. Therefore, in the whole process,

the participation of stakeholders and local people is also nec-essary and very important.

3.3. Goals and Decision-Making Forms of Rural TourismLandscape Governance. The protection, restoration, andstrengthening of rural tourism characteristics are notintended to be an obstacle to the creation of new rural tour-ism landscapes. When it is concluded that certain types ofrural tourism characteristics or areas are suitable forstrengthening or renewal strategies based on evaluationand analysis, this means that there is room for huge changesin such rural tourism landscapes. Similarly, many degradedand declining rural tourism landscapes also urgently needactive rural tourism landscape reconstruction to improveenvironmental quality and people’s quality of life, such asold industrial areas, wetlands, and swamp areas that needto be rebuilt and renewed. The analysis of rural tourismcharacteristics plays a key role in determining the rural tour-ism landscape areas with the potential for renewal and pro-motion. They can assist the restoration of the lost valuablerural tourism features and the investigation of the creationof new rural tourism landscapes.

The method of making judgments based on characteris-tics will vary with the transaction to be determined and thespecific situation that must be implemented. There must bea clear logic behind the method of goal judgment. It is nec-essary to carefully consider the overall characteristics andkey characteristics of the rural tourism landscape, its historyand origin, recent changes, changing trends, and future driv-ing forces. There may be the creation of new rural tourism.This will help realize rational judgments about futurechanges in existing features.

Development control also needs to integrate the charac-teristics of rural tourism. The purpose is to ensure that the

Evaluation of the characteristics of rural tourism landscape

Sensitivity and carrying capacity evaluation Fieldwork

Consult

Desktop researchLandscape sensitivity

Visual sensitivity

Historical and cultural sensitivity

Ecological sensitivity

Economic sensitivity

Landscape and visual sensitivity or integrated sensitivity research

Overall sensitivity

ValueBearing capacity

Alleviate or strengthen

Consult

Figure 2: Process diagram of rural tourism feature evaluation.


decision-making can combine the characteristics of ruraltourism as much as possible. The feasibility of the develop-ment project mainly depends on whether the developmentbehavior will have a negative impact on the rural tourismlandscape in the area. When relevant agencies adopt thismethod, they must formulate effective evaluation criteria(based on the key rural tourism characteristics and theimpact analysis of the development-sensitive rural tourismcharacteristics) to judge it. Development must meet thecharacteristics of the rural tourism policy objectives of theregion, the developer knows how to obtain the permit, andthe personnel of the relevant agencies know how to reviewthe development plan and finally form a high-quality devel-opment and construction.

The method of rural tourism landscape strategy is mainlyaimed at how the rural tourism landscape changes, using theassessment of rural tourism characteristics as the spatialframework to determine the development strategies (strength-ening, maintaining, restoring, etc.) of different regions. Thecoordination and judgment processes supporting the forma-tion of these strategies need to be clear and transparent, usu-ally involving subjects with different values and differentrural tourism landscape objectives, and are closely related tothe rural tourism landscape quality objectives. For example,in LEED’s rural tourism landscape strategy, different develop-ment strategies are determined based on the characteristicareas of rural tourism, such as protection, restoration orstrengthening strategies. It can be seen that different fromthe single protection method in the past, the document pro-vides more flexible strategies to guide different Regional devel-opment and changes. In the rural tourism landscape policydivision of Staffordshire, the development strategy is furtherupgraded to a policy division. The policy objectives includefive categories: innovative renewal, restoration, strengthening,maintenance, active protection, and areas characterized byrural tourism. The smallest unit of “Land Description Unit”(LDU) is the spatial framework to implement various policyobjectives into specific spatial scope, as the basis for futurejudgment and decision-making.

Environmental impact assessment or strategic environ-mental assessment for different types of development is animportant part of the development and construction assess-ment of the European Union and England. Rural tourismcharacteristics and visual impact assessment are the maincomponents of the research. The rural tourism landscapedepartment and the environmental assessment departmentfinally form an independent report. The general principlesand core of the method are similar in each case. The assess-ment of rural tourism landscape impact should define thetype of rural tourism landscape in the area where the devel-opment project is located. It analyzes how the various ele-ments that make up the rural tourism landscape, theaesthetic and perception aspects, the characteristics, andthe key characteristics that form the characteristics areaffected. It assesses the consequences (profits and losses) ofnot carrying out development and construction and assessesthe measures and scope of mitigating the impact. Rural tour-ism feature assessment can guide all stages of developmentand construction under the framework of environmental

impact assessment, so as to guide construction projects torespond better to the background of rural tourism landscape.

4. FP_Apriori Algorithm Based on FrequentPattern Tree FP-Tree

4.1. The Idea of FP_Apriori Algorithm. Through the analysisof the FP-Growth algorithm, a large transaction library canbe efficiently compressed using the frequent pattern treeFP-tree, which can completely avoid the huge overhead ofrepeatedly scanning the transaction library, which directlyand effectively solves the classic Apriori algorithm relyingheavily on the I/O overhead of the system.

In the traditional Apriori algorithm, we use Lk−1 self-connection to get the candidate item set Ck and then scanthe database to get the frequent item set Lk. In this process,there are two major shortcomings: the candidate item setCk is too bloated, which will produce a particularly largenumber of useless candidate item sets; second, when deter-mining the frequency of the candidate item set, the entiredatabase needs to be repeatedly traversed and scanned. Theoperating efficiency is very poor. We can also thin the fre-quent item set Lk−1 in advance according to the pruningstrategy, and optimize the frequency calculation process ofthe generated candidate item set. The specific optimizationstrategy will be studied in detail below.

Combining the above improvement ideas can effectivelysolve and overcome the shortcomings of the traditionalApriori algorithm, thereby further improving the executionefficiency and applicability of the algorithm, and can alsoserve as a reference for the improvement of other associationrule algorithms.

4.2. Frequent Pattern Tree FP-Tree. First, I give the relevantdefinition of the frequent pattern tree FP-tree.

Definition 1. Frequent pattern tree FP-tree is composed of aroot node (null), a group of root node item prefix subtrees,and an item header table.

Definition 2. All nodes of the item prefix subtree are com-posed of three parts: item name, node count, and nodechain. Among them, the count of the node represents thenumber of transactions that contain the node in the data-base, and the node chain points to the next node with thesame item name in the frequent pattern tree. If there is nonext node, it is empty. In particular, in the FP_Apriori algo-rithm, some improvements need to be made to FP-tree; thatis, all nodes of the item prefix subtree also need to record thename of the transaction that contains the node.

Definition 3. The item header table mainly contains twoattributes: the item name and the node chain head. The nodechain head is the node that points to the first item in the FP-tree.

I generate a bit matrix containing only 0 and 1 by scan-ning the transaction library. In this matrix, each row repre-sents a transaction Ti, and each column represents an item


set I j, where 0 and 1, respectively, indicate whether the itemset I j is in appears in this transaction Ti. Suppose the trans-action set is

T = T0 T1 T2 ⋯ Tm−2 Tm−1½ �: ð1Þ

The collection of all item sets is

I = I0 I1 I2 ⋯ In−2 In−1½ �: ð2Þ

The specific definition of the matrix is

M = Tij

� �m−1ð Þ⋅ n−1ð Þ

=Tij = −1, I j ∉ T ,iTij = 1, I j ∈ T ,

8<

:i = 0, 1, 2,⋯,m − 1, j = 0, 1, 2,⋯, n − 1:

ð3Þ

Insert a column into the matrix M to record the numberof transactions with the same item set, denoted as RC; ifthere are no duplicate transactions in the transaction data-base, set this value to 1. In this way, the order of the matrixis reduced very effectively. The specific definition of RC is

RCi =RCi − 1, Tsj = Ttj,−1, Tsj ≠ Ttj,

(

j = 0, 1, 2,⋯, n − 1, t = 0, 1, 2,⋯,m − 1:

ð4Þ

The support of all 1st-order item sets is the sum of non-zero elements in each column, which is specifically definedas

Sum = 〠m−1

i=0TijTij > 0, j = 0, 1, 2,⋯, n − 1: ð5Þ

The support of all k-th order item sets can be calculatedby RC and Tij, which is specifically defined as

Count k‐Item setð Þ = 〠n

i=1〠k

j=1Tij ⋅ RCi

� �: ð6Þ

According to the above definition, the specific algorithmfor constructing the FP-tree can be given as shown inFigure 3.

4.3. Optimization Strategy for Frequency Calculation ofCandidate Item Sets. In the improved algorithm FP_Aprioriin this article, first, you scan the entire transaction databaseto get the first-order frequent item set L1 and its correspond-ing support and transaction identification Tid and then con-tinuously iterate through L1 to generate L2, L3,…, Lk.

After self-connection through L1, the second-order can-didate set C2ðx, yÞ is obtained, where x and y are both itemsin the second-order candidate set C2. At this time, beforesearching the frequent pattern tree FP-tree to calculate thecandidate item set frequency, I obtain the item with less sup-

port between x and y according to L1. In this way, calculat-ing the candidate item set frequency only needs to scan thetransaction database. Specific transactions do not need toscan the entire database; in other words, we only need tosearch for specific branches of the FP-tree to calculate thefrequency of candidate item sets. Similarly, I generate C3ðx, y, zÞ through L2 self-connection; first, we find the item withthe least support of x, y, and z. At this time, you only need tosearch for the specific branch of the FP-tree to calculate theC3ðx, y, zÞ support degree.

Through the above optimization strategy, the supportdegree of the candidate item set can be calculated fasterand more directly, and to a certain extent, the execution effi-ciency of the algorithm can be greatly improved.

4.4. FP_Apriori Algorithm Analysis. The FP_Apriori algo-rithm is a comprehensive manifestation of the traditionalApriori algorithm and the frequent pattern tree FP-tree.Transplanting FP-tree to the Apriori algorithm can effec-tively improve the execution efficiency of the algorithm.The specific improvement work of FP_Apriori is mainlyreflected as follows:

(1) The FP_Apriori algorithm uses FP-tree to compressthe transaction database, which can specifically solvethe problems of the traditional algorithm scanningthe database too many times, the I/O burden beingtoo heavy, and the efficiency being too low

(2) It adopts the strategy of prethinning Lk−1 by usingthe Mapping_Apriori algorithm. The Ck generatedby the self-connection of Lk−1 of this strategy is anecessary condition for frequent item sets. The fre-quency of the first i items in any item set of Lk−1 isat least k − i. In the same way, the FP_Apriori algo-rithm can also thin L1 in advance to avoid the gener-ation of useless candidate items

5. Experiment and Analysis

5.1. Data Source. This paper collects network comment dataof 50 rural tourist spots as basic research data. Consideringthat the rural tourism market mainly comes from the localarea, and the O2O platform for local life is also an importantchannel for residents to purchase and evaluate leisure expe-rience products, this study added Dianping.com as a net-work data source website. Dianping.com includes small-and medium-sized rural tourist spots that are not coveredby other large-scale tourism websites. The emphasis of theword-of-mouth content of each website is different, whichcan ensure the richness and completeness of the informationsource. The information comparison of the five text sourcewebsites is shown in Figure 4.

5.2. Data Characteristics. I read through all the samples toscreen the samples one by one and delete duplicate and irrel-evant information. The word-of-mouth evaluation needs toinclude the comprehensive experience evaluation of the pro-ject; that is, it needs to include the tourist’s travel itinerary inrural tourism, the same tourist group, service quality,


environmental quality, accommodation conditions, trans-portation accessibility, product evaluation, willingness torevisit, etc. After screening and cleaning, there are 19,916remaining. Flower port and urban vegetable garden in a cer-tain area are national 4A-level scenic spots. The distributionof online text data among scenic spots indirectly supports

the problem of market concentration in the rural tourismindustry, that is, the high market concentration of farmproducts. Although there are a large number of rural tour-ism projects in a certain area, the well-known farms (villas)are the main force for receiving rural tourists. The numberof reviews in each month is shown in Figure 5.

Start

Input transaction library

Set minimum support

The first traversal scans the transaction database D

Find the 1st order frequent itemset L1 and the degree of support

Sort in descending order of support

Record the result as L

Create the root node of the frequent pattern tree FP-tree

Mark with null

Sort transactions according to L

Recursive call

Does transaction T have child

nodes N ?

Then the count of N increases by 1

Yes

Recreate a new node N

Record its count as 1for the first time

Connect it to its parent node

Output frequent pattern tree FP-tree

End

Get sorted frequent items

No

Figure 3: Algorithm for generating frequent pattern tree FP-tree.

1 2 3 4 5 6 7 8 9 10 11 1262

64

66

68

70

72

74

76

78

80

82

Month

Scen

ic co

vera

ge (%

)

CtripDonkey mother netDianping

QunarTongcheng

Figure 4: Comparison of text source website information.


5.3. Analysis of Market Demand Heat. Network text analysismethods are used more and more widely in tourismresearch, especially the research on the image of tourist des-tinations and the experience of tourists. Collecting onlinetravel word-of-mouth from multiple destinations for textanalysis and cross-comparative analysis can not only refinethe market demand characteristics of tourists but also studythe supply differences between destinations and find the gapbetween destination supply and demand. This research willuse about 20,000 online review data from 50 major ruraltourist spots in a certain area on 5 major online platformsand use text analysis methods, statistical testing methods,cluster analysis methods, common network analysis, and

other statistical methods for analysis. We combine theresearch on the characteristics of rural tourism demand ina certain area with the characteristics of tourism supply ofrural tourism projects to find the direction of the

1 2 3 4 5 6 7 8 9 10 11 122000

2500

3000

3500

4000

4500

5000

Month

Num

ber o

f rev

iew

s

DianpingQunarDonkey mother net

TongchengCtrip

Figure 5: Distribution of the number of text data before filtering.

1 2 3 4 5 6 7 8 9 10 11 121800

2400

3000

3600

4200

4800

5400

6000

6600

7200

Month

Rura

l tou

rist a

rriv

als

CtripTongcheng

Donkey mother netQunar

Figure 6: Monthly distribution of rural tourism tourist volume data.

Table 1: Test of the relationship between demand heat andwhether the scenic spot has won the title.

t test for the mean equation

t Df Sig.Mean

differenceStandarderror

95% lowerlimit

95% upperlimit

1.66 43 0.08 28 18 2 64

2.21 46.3 0.03 28 13 5 55


optimization of the supply and demand structure of ruraltourism in a certain area.

Although there is a time lag between the travel time andthe review time of tourists, the time distribution of the net-work review data of scenic spots can basically reflect thetravel situation of the market in different time periods. Thenumber of online reviews of major rural tourism projectsin a certain area has two obvious peak periods, and thereare differences in geographical space, reflecting the seasonalcycle of the industry. As shown in Figure 6, September of theyear is the peak periods for online reviews. After calculation,the monthly average value of the number of reviews in sce-nic spots is 1,660, and the coefficient of variation is 0.204.The number of reviews in September has greatly exceededthe average. Based on the specific comments, it can be seenthat there is a concentrated period of outings for residentsin a certain area, flower viewing during festivals, and citizensgoing to the suburbs to sweep graves. The number of reviewsin the remaining months is more evenly distributed, which isthe result of multiple factors such as the repetitiveness ofrural tourism demand, the convenience of urban transporta-tion, and the convenience of online platforms. A stable pas-senger flow helps to maintain the balance of rural tourismsupply.

As branded rural tourism projects have a high degree ofpopularity, I will further examine the relationship betweenthe popularity of reviews and whether scenic spots havereceived relevant titles, that is, the relationship between thepopularity of scenic spots and whether they are scenic spotsabove 3A or demonstration sites for leisure agriculture andrural tourism. I mark the scenic spots that belong to at leastone of these types as “1,” and mark them as “0” if they areneither, and perform an independent sample t test. The testresults are shown in Table 1.

Levene’s test confirms that the variances are homoge-neous, and “assuming the variances are equal” is used. The

P value of the two-tailed significance level of the indepen-dent sample t test is 0.094, which is significant at the 10%level. The test results show that the average monthly onlinereviews of scenic spots above 3A and demonstration sitesfor leisure agriculture and rural tourism are higher thanthose of rural tourism projects that have not obtained theabove titles. These projects have a relatively high level ofreception, which also confirms the market utility and brandutility of various standardized demonstration sites.

5.4. Demand Experience Feature Extraction Experiment. Byextracting the basic commonality of tourism demand andexperience of each project, the demand attribute of ruraltourism industry is analyzed. The word-of-mouth docu-ments of each scenic spot are processed one by one, andthe software is used to extract the high-frequency featurewords of the text documents of each scenic spot. In orderto ensure the completeness of the travel experience summaryand the accuracy of the description, according to the attri-bute classification of nouns, verbs, and adjectives, the topten words of the three types of words in the document areextracted, totaling 30 words. I check the high-frequency fea-ture words extracted from the text research files of each sce-nic spot one by one, remove words that do not containtourism experience and other tourism characterizationmeanings, remove the vocabulary that characterizes the sce-nic spot name and the area, and supplement it with subse-quent vocabulary. According to the frequency of thevocabulary of all 30 high-frequency words in each scenicspot, they are extracted at least sequentially until each scenicspot successfully extracts 20 high-frequency feature words.Four scenic spots were removed due to the insufficient num-ber of comments and the inaccurate extraction of high-frequency words.

I sort out the vocabulary of the high-frequency featurewords in the network review data text of the scenic spot,

10 20 30 40 50 60 70 80 905

10

15

20

25

30

35

40

99 characteristic words

Freq

uenc

y

DianpingCtripDonkey mother net

QunarTongcheng

Figure 7: Common feature extraction results of rural tourism experience.


so as to extract the common characteristics of the touristdemand of the scenic spot. There are 99 common character-istic words in rural tourism experience, as shown in Figure 7.The characteristics of rural tourism tourist experience in acertain area and the overall characteristics of tourist demandrevealed from it are basically consistent with the characteris-tics of rural tourism demand. It is also explained from theside that the text analysis method used to evaluate thedemand attributes of rural tourism reflects high accuracyand reliability.

5.5. Market Satisfaction and Overall Evaluation. I analyzetourists’ overall satisfaction with the scenic spot based onthe proportion of positive emotions in text documents andfurther understand the structural gap between industry sup-ply and demand. In order to comprehensively weigh theoverall satisfaction of tourists with rural tourism productsin a certain area, first, the sentiment analysis method is usedto calculate the proportion of positive emotions and negativeemotions in the network text data of scenic spots. The posi-tive emotions in most of the tourist spots’ online review datatexts are more than 70%, indicating that tourists’ overall sat-isfaction with the rural tourism experience is relatively high.Considering the small number of words in a single onlinereview data, about 30% of the negative sentiment proportionin the online review data also indicates that there is stillroom for improvement in the product quality and servicequality of rural tourism in a certain area. Due to the largedifferences in the amount of network data information indifferent scenic spots, in order to more objectively reflectthe differentiated characteristics of different scenic spots,the proportions of positive emotions in Internet word-of-mouth with different feature words are compared here. Asshown in Figure 8, the FP_Apriori algorithm performs betterthan the Apriori algorithm in the proportion of positiveemotions.

6. Conclusion

This paper improves an improved Apriori algorithm basedon the frequent pattern tree FP-tree. The idea of thisimproved algorithm is to combine the traditional Apriorialgorithm with FP-tree, which can operate on transactions,compress them efficiently, and avoid a large number ofrepeated transaction database traversals and scans. A newtransaction database scanning strategy is proposed, whichcan scan only specific databases, making the scanning morepurposeful and pertinent, and avoiding a lot of useless oper-ations. The empirical research concludes that basic tourismconsumption such as catering is dominant as a rural tourismdemand motive, and demand motives such as leisure vaca-tion and in-depth experience are insufficient. Touristsdescribe the experience content more than their spiritualperception, reflecting the slower and slower demand escala-tion. The overall satisfaction of rural tourists is relativelyhigh, but they are also more dissatisfied with the level of ser-vice, the level of reception facilities, and tourism informationservices. Generally speaking, there are still big problems inthe visit rate, popularity, and reputation of most rural touristdestinations in a certain area. In addition, the use of textanalysis methods to extract rural tourism needs and experi-ence characteristics has high accuracy. The travel motives,consumption characteristics, and organizational characteris-tics of rural tourism demand in a certain area extracted bythe text analysis method are basically consistent with therural tourism demand characteristics summarized in the lit-erature review. The above empirical analysis reveals thestructural contradictions within and between the two mainbodies of rural tourism supply and demand. The supply ofrural tourism urgently needs to innovate products and ser-vices to provide tourists with rural tourism experience prod-ucts that match the quality and price and can lead the trend.The optimization of the rural tourism market structure also

10 20 30 40 50 60 70 80 9086

88

90

92

94

96

99 characteristic words

Perc

enta

ge o

f pos

itive

emot

ions

(%)

Apriori algorithmFP-Apriori algorithm

Figure 8: Proportion of positive emotions in the network data of rural tourist attractions.


requires the establishment of a benign interaction mecha-nism between industrial supply and industrial demand, lead-ing demand upgrading through supply innovation, andpromoting supply renewal through demand upgrading,forming a continuous driving force for the overall progressof the industry.

Data Availability

The data used to support the findings of this study areincluded within the article.

Conflicts of Interest

The author does not have any possible conflicts of interest.

References

[1] K. Pitchayadejanant and P. Nakpathom, “Data miningapproach for arranging and clustering the agro-tourism activ-ities in orchard,” Kasetsart Journal of Social Sciences, vol. 39,no. 3, pp. 407–413, 2018.

[2] F. Jiang, K. K. R. Yuen, and E. W. M. Lee, “Analysis of motor-cycle accidents using association rule mining-based frame-work with parameter optimization and GIS technology,”Journal of Safety Research, vol. 75, pp. 292–309, 2020.

[3] I. Khatri, “Information technology in tourism & hospitalityindustry: a review of ten years publications,” Journal of Tour-ism and Hospitality Education, vol. 9, pp. 74–87, 2019.

[4] Y. Li, H. Zhang, D. Zhang, and R. Abrahams, “Mediatingurban transition through rural tourism,” Annals of TourismResearch, vol. 75, pp. 152–164, 2019.

[5] N. Cucari, E. Wankowicz, and S. E. De Falco, “Rural tourismand Albergo Diffuso: a case study for sustainable land-useplanning,” Land Use Policy, vol. 82, pp. 105–119, 2019.

[6] W. Yu, “Discovering frequent movement paths from taxi tra-jectory data using spatially embedded networks and associa-tion rules,” IEEE Transactions on Intelligent TransportationSystems, vol. 20, no. 3, pp. 855–866, 2019.

[7] F. Xu, N. Nash, and L. Whitmarsh, “Big data or small data? Amethodological review of sustainable tourism,” Journal of Sus-tainable Tourism, vol. 28, no. 2, pp. 144–163, 2020.

[8] P. Siakwah, R. Musavengane, and L. Leonard, “Tourism gover-nance and attainment of the sustainable development goals inAfrica,” Tourism Planning & Development, vol. 17, no. 4,pp. 355–383, 2020.

[9] M. M. Su, G. Wall, and S. Wang, “Yujiale fishing tourism andisland development in Changshan archipelago, Changdao,China,” Island Studies Journal, vol. 12, no. 2, pp. 127–142,2017.

[10] Q. He, W. He, Y. Song, J. Wu, C. Yin, and Y. Mou, “The impactof urban growth patterns on urban vitality in newly built-upareas based on an association rules analysis using geographical'big data',” Land Use Policy, vol. 78, pp. 726–738, 2018.

[11] H. Q. Vu, G. Li, R. Law, and Y. Zhang, “Tourist activity anal-ysis by leveraging mobile social media data,” Journal of TravelResearch, vol. 57, no. 7, pp. 883–898, 2018.

[12] W. Wang, J. Wu, M. Y. Wu, and P. L. Pearce, “Shaping tour-ists’ green behavior: the hosts’ efforts at rural Chinese B&Bs,”Journal of Destination Marketing & Management, vol. 9,pp. 194–203, 2018.

[13] R. Manikandan and V. Saravanan, “A novel approach on par-ticle agent swarm optimization (PASO) in semantic mining forweb page recommender system of multimedia data: a healthcare perspective,” Multimedia Tools and Applications, vol. 79,no. 5-6, pp. 3807–3829, 2020.

[14] I. O. Ezeuduji, “Change management for sub-Saharan Africa'srural tourism development,” Current Issues in Tourism,vol. 20, no. 9, pp. 946–959, 2017.

[15] C. H. Chin, M. C. Lo, A. A. Mohamad, and V. Nair, “Theimpacts of multi-environmental constructs on tourism desti-nation competitiveness: local residents’ perceptions,” Journalof Sustainable Development, vol. 10, no. 3, pp. 120–132, 2017.

[16] F. Wang, L. Xia, Z. Chen, W. Cui, Z. Liu, and C. Pan, “Remotesensing identification of coastal zone mariculture modes basedon association-rules object-oriented method,” Transactions ofthe Chinese Society of Agricultural Engineering, vol. 34,no. 12, pp. 210–217, 2018.

[17] X. Zhao, X. Lu, Y. Liu, J. Lin, and J. An, “Tourist movementpatterns understanding from the perspective of travel partysize using mobile tracking data: a case study of Xi'an, China,”Tourism Management, vol. 69, pp. 368–383, 2018.

[18] J. Yuan, H. Lin, and X. Li, “Analysis and comments on thecourse of rural development,” Agricultural Science & Technol-ogy, vol. 18, no. 12, pp. 2653–2657, 2017.

[19] K. Güler and Y. Kâhya, “Developing an approach for conserva-tion of abandoned rural settlements in Turkey,” A/Z ITU Jour-nal of the Faculty of Architecture, vol. 16, no. 1, pp. 97–115,2019.

[20] T. C. Leu, M. Eriksson, and D. K. Müller, “More than just a job:exploring the meanings of tourism work among indigenousSámi tourist entrepreneurs,” Journal of Sustainable Tourism,vol. 26, no. 8, pp. 1468–1482, 2018.

[21] X. Yang and P. Ho, “Mining institutions, contention and cred-ibility: applying the conflict analysis model to court cases inChina,” The Extractive Industries and Society, vol. 7, no. 3,pp. 1011–1021, 2020.

[22] R. Armis and H. Kanegae, “The attractiveness of a post-miningcity as a tourist destination from the perspective of visitors: astudy of Sawahlunto old coal mining town in Indonesia,”Asia-Pacific Journal of Regional Science, vol. 4, no. 2,pp. 443–461, 2020.


analysis of rural tourism demand characteristics and

Documents