prediction of construction dispute using artificial...
TRANSCRIPT
http://www.iaeme.com/IJCIET/index.asp 582 [email protected]
International Journal of Civil Engineering and Technology (IJCIET) Volume 10, Issue 01, January 2019, pp. 582-594, Article ID: IJCIET_10_01_054
Available online at http://www.iaeme.com/ijciet/issues.asp?JType=IJCIET&VType=10&IType=01
ISSN Print: 0976-6308 and ISSN Online: 0976-6316
© IAEME Publication Scopus Indexed
PREDICTION OF CONSTRUCTION
DISPUTE USING ARTIFICIAL NEURAL
NETWORKS TESTIMONIES FROM
INDIAN CONSTRUCTION PROJECTS
Asra Fatima
Research scholar, GITAM (Deemed to be University), Hyderabad, India.
Dr. Bellam Sivarama Krishna Prasad
Head & Professor GITAM (Deemed to be University), Hyderabad, India.
Dr. T.Seshadri Sekhar
Dean& Professor, NICMAR, Hyderabad, India
ABSTRACT
The purpose of this paper is to facilitate construction project team members to
figure out the factors that they must firmly monitor to complete projects without any
dispute and also to predict construction dispute attributes during the course of a
project. The study conferred here is an extension of past research in which 53
project dispute attributes were determined based on expert’s opinion and literature
survey which after analysis resulted in 4 factors which have been used to develop
the dispute attribute prediction model based on artificial neural networks (ANN).
The analyses of the responses led to the conclusion that factors such as Time
phasing and requisite contracting legislation, Project financials and client contractor
partnering, Quality and risk management under ambiguity, Non-responsive owner and
unrealistic contractor rules significantly affect the project dispute attributes. The
best prediction model was found to be a 5-9-1 feed forward neural network based on
back propagation algorithm with a mean absolute percentage deviation (MAPD) of
12.22 percent.
Keywords: Project dispute attributes, Construction works, India, Neural nets.
Cite this Article: Asra Fatima, Dr. Bellam Sivarama Krishna Prasad and Dr.
T.Seshadri Sekhar, Prediction of Construction Dispute Using Artificial Neural
Networks Testimonies From Indian Construction Projects. International Journal of
Civil Engineering and Technology, 10(01), 2019, pp. 582–594
http://www.iaeme.com/IJCIET/issues.asp?JType=IJCIET&VType=10&IType=01
Asra Fatima, Dr. Bellam Sivarama Krishna Prasad and Dr. T.Seshadri Sekhar
http://www.iaeme.com/IJCIET/index.asp 583 [email protected]
1. INTRODUCTION
Management of engineering of construction projects is a complex affair (Anderson, 1992). The
participation of the owners, designers, contractors, sub-contractors, specialists, consultants etc.
define the multidisciplinary outlook of the constructions projects. With the increase in extent
of the project, the number of participants in the project is also compelled to increase which
leads to dispute.
The construction sector asserts to account for 11 percent of India’s GDP and is predicted
to be on the rise. Infrastructure, airport, metro rail and power sector projects constitute an
important role in this sector. Cost and time are considered as important elements in most
construction projects. The objectives of the study are set as follows:
• Identify the factors that can be used to predict the project dispute attributes of a
BOT project; and
• Develop a model to predict the project dispute attributes of a BOT project using
artificial neural network (ANN).
2. LITERATURE REVIEW
Some of the past researchers have taken up questionnaire survey approach for collection of
data and employed mathematical tools like AHP, Neural networks and statistical techniques
like factor analysis and multivariate regression, etc., for analysis and drawing conclusions.
Summary of studies in the field of project dispute factors are as follows,
Willaim H.Ross Jr (1988) stated that a survey was conducted to investigate the beliefs of
alternative dispute resolution (ADR) professional about the impact of various situational
factors on the likelihood of negotiating an agreement.
Charles et al(1990) stated that the results of an investigation that was conducted to assess
professional opinion regarding the course of construction disputes, the types of owners most
often involved in disputes, the attorney's role in dispute resolution and the keys for avoiding
disputes.
S. O. Cheun et al (2000) stated that an artificial neural network analysis helps in
determining the important factors affecting the construction dispute resolution processes in
Hong Kong.
K.N. Jha et al (2009) described that project manager always encounters difficulties in
predicting the performance of a construction project.
Ibrahim Yitmen & Ebrahim Soujeri (2010) stated that the purpose of this paper was to
identify the negative impact of change orders on construction productivity, which leads to
decrease in labor efficiency, loss of man hours and considers an ANN Model which estimates
the probable disputes which may be avoided or resolved before litigation occurs.
N. B. Chaphalkar et al (2015) stated that the use of artificial neural networks can be a
cost-effective method to help to forecast the outcome of construction claims and disputes.
3. RESEARCH METHODOLOGY
The methodologies to achieve the objectives of identification of factors which can be used to
predict the dispute attributes of BOT projects and development of prediction model for the
same have been dealt with in this section. This study identified the relative importance of
project dispute attributes in Indian construction industry and also found the effect of these
Factors BOT project. The analyses of the first stage questionnaire survey resulted in the
exclusion of the 53 project dispute attributes identified based on experts opinion and literature
Prediction of Construction Dispute Using Artificial Neural Networks Testimonies From Indian
Construction Projects
http://www.iaeme.com/IJCIET/index.asp 584 [email protected]
survey. The analyses resulted in a number of common factors among them, while a few other
factors emerged predominantly in only one or other criterion. The union of all common and
uncommon factors across two criteria resulted in18 factors. These factors are reproduced in
Table I.. In the survey, the respondents were asked to consider a ‘‘choice’’ project, and were
asked to give the extent of contribution of the 20 factors in a 5-point scale (1-5, with 1
indicating high negative contribution, 0 being no effect and 5 indicating high positive
contribution) on the project dispute of the project. A total of 30 responses were received out of
the150 mailed for the second stage questionnaire survey. With the 18 factors as explanatory
variables and the contribution of these factors on the arousal of dispute as the response variable,
multiple regression was carried out in the study and the conclusions were drawn. Thus, there
is a need to identify the predictor variables among the 18 factors and develop a project dispute
forecasting model.
Table 1.
Table 1.Table 1.
Table 1. Factors affecting dispute in BOT project
Factors Description of factors
F1 Technical inadequacy of the contractor
F2 Arability of Resources
F3 Lack of Construction legislation
F4 Payments Delays
F5 Work availability
F6 Client experience
F7 Escalation payable to the contractor as per the Provision in contract
F8 Escalation due to price variation during contract period
F9 Non retention of refund money
F10 Previous working
F11 Client Budget
F12 Relationship of contractor with Prime contractor
F13 Ambiguities in contract documents
F14 Risk Allocation
F15 Quality of Work
F16 Failure respond in timely manner
F17 Lack of information from design team /Client
F18 Restriction Place don contractor working method
Asra Fatima, Dr. Bellam Sivarama Krishna Prasad and Dr. T.Seshadri Sekhar
http://www.iaeme.com/IJCIET/index.asp 585 [email protected]
Figure 1. Research methodology as a flow chart.
4. PREDICTION MODEL
To ensure that these factors are the key determinants that affect project dispute
and also to develop the dispute prediction model for the same, ANN methodology
was adopted. ANN has been adopted to consider the factors that alter the different aspects of
DB project performance in Singapore construction projects (Ling et al., 2004a). It is actually a
mathematical model in which the information transformation take place in a number of simple
elements called neurons (nodes), and the signals are transmitted between neurons over
Prediction of Construction Dispute Using Artificial Neural Networks Testimonies From Indian
Construction Projects
http://www.iaeme.com/IJCIET/index.asp 586 [email protected]
connection links that has a correlated weight with it. Each neuron donate an activation function
(transfer function) to the incoming signal to determine its output signal (Zurada,
1992).McCulloch and Pitts chose a binary threshold unit as a computational modelfor an
artificial neuron as shown in Figure 2. This mathematical model calculate aweighted sum of its
n input signals Xj = (1, 2. . . n) and developed an output of 1, if it is above a certain threshold
or else give an output of 0 (referred in Zurada, 1992).The ANN models choose various
activation functions such as linear, sigmoid, andGaussian functions out of which the sigmoid
is the most commonly used function (Jainetal.1996).To specify mathematically, each neuron j
sums its weighted input as mentioned below:
The output of a neuron, y is a function of its weighted input, mentioned as follows:
Figure 2. MuCuulloch-Pitts model of a neuron
4.1. Network Structure
Feed forward neural network architecture has been chosen for constructing the ANN model. It
consists of an input layer, output layer and hidden layers if required. The input layer provide
input data to the network, and the extent of the input layer, or the number of neurons (nodes)
Which are associated factors for the project dispute criterion in this case. When the hidden
layers are used, the number of nodes in the hidden layer is determined by trial and error. The
hidden nodes on accepting the values of the inputs compute the weighted sum of the inputs and
according to the transfer function compress the values to a limited range (Edwards, 2007).
Table II gives the information of the sigmoid and linear transfer functions used in MATLAB
software.
4.2. Training model
The main goal of neural network training is to decrease the output error by regulating network
weights and biases. Many learning algorithms have been developed for ANN. According to
Jain et al. (1996) back propagation learning algorithm with feed forward network architecture
is most appropriate for predictions. Back propagation is a learning algorithm for a feed forward
network which is an administrated learning process based on error correction learning rule. In
back propagation input and the corresponding output are used to train a network continuously
so that it can approximate a function and correlate input with specific output. Properly trained
back propagation networks provide reasonable answer when given with new inputs in most
cases. The inputs are shipped forward and then the errors are propagated backwards. The
training generally starts with random weights and biases, which are adjusted by the algorithm
for decreasing errors. The regular back propagation in MATLAB is gradient descent algorithm.
Asra Fatima, Dr. Bellam Sivarama Krishna Prasad and Dr. T.Seshadri Sekhar
http://www.iaeme.com/IJCIET/index.asp 587 [email protected]
Other valuable algorithms in MATLAB are Leven berg-Marquardt, Conjugate gradient
algorithm, resilient back propagation algorithm etc. In general no single algorithm matches all
applications, the constants strategy is to examine with these algorithms and find the most
suitable one for the given application.
4.3. COMPOSITION OF ANN
Once the above particulars are defined, the input–output data is divided towards training and
validation data sets. There are two methods applicable for training and testing the models – the
hold out method and a re-sampling method called random sub-sampling (Edwards, 2007). In
case of the hold out method, the data sets are split into two–usually 2/3 for designing (training
set) and 1/3 for the reckoning the true performance(testing and validation set). While in the
latest, various random train and test experiments are performed on training and testing samples.
And according to Edwards (2007) and Goh (1995) when the number of data sets is adequate
the hold out method proves to be effective. Hence, about two-thirds of the data are used for
training and the remaining for testing and validation. As per the procedure mentioned in Figure
1, the network is trained with training data set, which is halted once the mean square error
(MSE) is satisfactory, and for this trained network the validation process is carried out. If the
mean absolute percentage deviation (MAPD) is found to be satisfactory it is treated as the
validated prediction model.
Table 2. Description of transfer function
4.3. Validation
The objective of operating validation was to test the competence of this trained network for
assessing unknown data sets. Predicted dispute attributes measures derived from the models
are compared with actual dispute attributes of projects and Table3givesthedispute attributes
measures used to validate the prediction models. In Table 3, ‘‘n’’ denotes for the number of
predictions. When the results of the validation are found to be satisfactory, then the model
would be expected to perform well when exposed to new data sets.
4.4. Data analysis
In this section correlation analysis and training and validation of developed ANN models for
prediction of dispute attributes have been discussed. As discussed earlier, 30 data sets received
from the second stage questionnaire survey were available for analysis.
Prediction of Construction Dispute Using Artificial Neural Networks Testimonies From Indian
Construction Projects
http://www.iaeme.com/IJCIET/index.asp 588 [email protected]
Table 3. Measures to validate the prediction models
4.5. Dispute prediction model.
Among the 72 data sets, 11 user’s data sets were considered for training and the left over 11
data sets were utilized for validation. MATLAB 17 software was used to design and train the
ANN models. A feed forward neural network based on back propagation was considered in the
ANN model training and sigmoid transfer function was used for the nodes. Training algorithm
and transfer functions were chose based on trial and error procedures.
The Levenberg-Marquardt back propagation algorithm (trainlm in MATLAB) and the
transfer function especially hyperbolic tangent function (tansig in MATLAB) for the neurons
in the hidden layers was found to give brisk convergence and better results amidst training and
validation. Summary of the dispute measures such as MAPD, MSE and root mean square error
(RMSE) for the models derived by differing the number of hidden layers and the number of
neurons in each layer is given in Table 4.
Based on the MAPD of the validation data, the 5-9-1 structure has the lowest MAPD among
all others in single hidden layer network structure, The MAPD for the 5-9-1 model developed
to predict dispute occurrence is found to be 12.22417 percent and emerge to be relatively robust
and the RMSE for the same is found to be ± 0.99 percent, which is acceptable (Ling et al.,
2004a). The training curve of 5-9-1 model for ‘‘dispute attributes’’ is shown in Figure3.
Figure 3. Training Curve of ANN “5-9-1project dispute attribute Model”
Asra Fatima, Dr. Bellam Sivarama Krishna Prasad and Dr. T.Seshadri Sekhar
http://www.iaeme.com/IJCIET/index.asp 589 [email protected]
Table 4. Summary of ANN model network structure
S
no
Number of
hidden
layer
Network
structure Training
cycle MSE
(Train) MAPE (%)
MSE
(Validation) RMSE
(Validation)
1 1 4-1-1 1000 0.25 27.8597 0.762129 0.873
2 2 4-2-1 1000 0.1936 18.4192 0.7921 0.89
3 3 4-3-1 100 0.1024 17.59896 0.7744 0.88
4 4 4-4-1 1000 0.4356 17.19583 0.935089 0.967
5 5 4-5-1 100 0.5625 19.19278 0.8464 0.92
6 6 4-6-1 1000 0.4624 20.96509 0.9604 0.98
7 7 4-7-1 100 0.3844 18.75861 0.81 0.9
8 8 4-8-1 100 0.64 17.11852 0.9604 0.98
9 9 4-9-1 100 0.81 12.22417 0.9801 0.99
10 10 4-10-1 500 0.7569 17.58769 0.9604 0.98
Table 5. Comparison between Actual and predicted for 5-9-1 ANN network
Sno Actual prediction ANN 5-9-1 prediction Absolute (PD) % 1 2 1.9758 1.21 2 4 3.9612 0.97 3 4 4.0005 0.0125 4 4 4.1631 4.0775 5 5 4.9133 1.734 6 4 4.0601 1.5025 7 3 3.0385 1.283333 8 4 4.0899 2.2475 9 5 4.979 0.42
10 3 3.0187 0.623333 11 1 1.1072 10.72 12 3 3.0185 0.616667 13 4 4.0385 0.9625 14 4 3.4461 13.8475 15 4 3.989 0.275 16 5 4.2825 14.35 17 3 3.9527 31.75667 18 5 3.9527 20.946 19 5 4.8862 2.276 20 5 4.8862 2.276 21 5 4.8862 2.276
The factors identified critical for project dispute are Time phasing and requisite contracting
Legislation, Project financials and client contractor partnering, Quality and risk management
under ambiguity, Non-responsive owner and unrealistic contractor rules the developed ANN
models had a feed forward network based on back propagation algorithm and the 5-9-1
structure gave the least MAPD of 12.22 percent considering the single hidden layer models.
Table VI gives the comparison of the actual and predicted performance for the 5-9-1 mode
Prediction of Construction Dispute Using Artificial Neural Networks Testimonies From Indian
Construction Projects
http://www.iaeme.com/IJCIET/index.asp 590 [email protected]
5. VALIDATION OF FINDINGS THROUGH CASE STUDIES AND
SECOND QUESTIONNAIRE
Details of Case Study: This research work deals with 3 case studies namely NH 44, NH65,
and NH163.
Description of NH 44 Case Study: National Highway 44 (NH 44) is the longest-running
major north–south National Highway in India. It starts from Srinagar and terminates in
Kanyakumari. NH-44 was laid and is maintained by Central Public Works Department This
highway starts from Srinagar and connects several cities and it is officially listed as running
over 3,745 km (2,327 mi) from Srinagar to Kanyakumari. It is the longest national highway
in India. The present study of NH 44 deals with the following route which is handled by
National Highways Authority of India (Project Implementing Unit (PIU) (Nirmal)). It started
on 2-11-2017 and planned to complete on 26-09-2009. There were 6 contractor associated with
the work and the following routes are given below.
Route -1 : It starts from 175.000 Km (Maharastra / AP Border) to Km. 230.000 (Islam
Nagar) on NH 44 in the State of Telangana under North – South Corridor (NHDP Phase II) on
Build, Operate and transfer (Annuity) Basis – Contract Package No. NS2/BOT/AP-6, Km.
230.000 (Islam Nagar) to Km. 278.000 (Kadthal) on NH 44 in the State of Telangana under
North – South Corridor (NHDP Phase II) on Build, Operate and transfer (Annuity) Basis –
Contract Package No. NS2/BOT/AP-7
Route -2: It starts from 230.000 Km. (Islam Nagar) to Km. 278.000 (Kadthal) on NH 44
in the State of Telangana under North – South Corridor (NHDP Phase II) on Build, Operate
and transfer (Annuity) Basis – Contract Package No. NS2/BOT/AP-7
Route-3: It starts from 278.000 Km. (Kadthal) to Km. 308.000 (Armur) on NH-7 in the
State of Andhra Pradesh under North – South Corridor (NHDP Phase II) on Build, Operate and
transfer (Annuity) Basis – Contract Package No. NS2/BOT/AP-8
Route -4: It starts from four laning of Armur – Adloor – Yellareddy section on Nagpur –
Hyderabad of NH 44 from Km 308.000 to Km 367.000 in the State of Telangana on DBFOT
Basis (Package No. NS2/BOT/AP-1)
Route -5: It starts from 4 laning from 367.000 Km to (Adloor Yellareddy) to Km 447.000
& Improvement, Operation & Maintenance from Km 447.000 to Km 464.000 (Gundla
Pochamapalli) of Nagpur – Hyderabad section of NH 44 in the State of Telangana (Package
No. NS-2/BOT/AP-2)
Route -6: It starts from Construction of four laning Km. 464.00 to Km. 474.00 Nagpur-
Hyderabad (NS-23). The following contractors dealt with the following routes and below are
their project cost.
Table 6. Details of list of contractors their respective routes and cost
Sno Name of the
stretch
List of Contractors handling the project as
per the route enlisted above
Project cost(Rs)
1 Route-1
Adilabad Expressway Pvt. Ltd. Rs. 360.42 Cr.
2 Route-2 Patel KNR Heavy Infrastructures Pvt. Ltd. Rs. 518.46 Cr.
3 Route-3 Nirmal BOT Ltd Rs. 271.73 Cr.
4 Route-4 Navayuga Dichpally Tollway Pvt. Ltd. Rs. 490.50 Cr.
5 Route-5 GMR Pochanpalli Expressway Pvt. Ltd Rs. 546.15 Cr.
6 Route-6 You-One Maharia (JV) Rs. 74.89 Crs
Asra Fatima, Dr. Bellam Sivarama Krishna Prasad and Dr. T.Seshadri Sekhar
http://www.iaeme.com/IJCIET/index.asp 591 [email protected]
The old name of highway was NH 7 connecting Varanasi to kanyakumari passing through
Uttar Pradesh, Madhaya Pradesh, Maharastra, Telangana, Andhra Pradesh, Karkataka and
Tamil nadu. The present study deals with the highway details of telangana state. There was an
increase in the traffic flows which led to congestions and reported more accident cases since
1998 to 2006, so the government took the decision of widening the national highway 7 from 2
lane to 6 lanes which was the renamed to NH44. This BOT project had many claims because
of the effect of identified factors they are as follows.
Adilabad Expressway Pvt. Ltd. commenced the work on 02.11.2007 during the period of
execution they claimed 48, 06, 35,404.00 rs against NHAI due to changes in design, idling of
plant and resources and their loss of overhead and fixed costs during the period of idling of
resources, which falls in the category of Non responsive owner and time phasing and requisite
contractor legislation (4th identified factor). Patel KNR Heavy Infrastructures Pvt. Ltd
commenced the work on 02.03.2008 during the period of execution they claimed
1,27,89,65,505.00 against NHAI due to Compensation For Non-Handing Over Of Land and
Claims For Arbitration Cost. which falls in the category Non responsive owner and time
phasing and requisite contractor legislation(4th identified factor).GMR Pochanpalli
Expressway Pvt. Ltd started the work on 26.09.2006 during the execution period they claimed
1152642326 rs against NHAI due to project financial and client contractor partnering and
quality and risk management under ambiguity(3rd identified factor).. You One - Maharia (JV)
commenced the work on 30.8.2001 during the execution period they claimed 61, 89, 45,600.00
rs due to Non responsive owner and unrealistic contractor rules and 3,42,00,44,032.00 rs for
project financial and client contractor partnering and quality. (2nd and 4th identified factor).
For validating the above case study details sample response of second questionnaire from
the companies is also considered and shown below and Spearman correlation is also performed
which is used to identify the factors that have high degree of association with the project dispute
factors using Statistical package for social sciences (SPSS 23.0). Therefore the 4 identified
factors were the key factors responsible for delays and claims for NH44.
5.1. Description of NH 163 Case Study
Four Laning of Hyderabad –Yadgiri section of NH 163(Old NH 202) from Km 18.600 to Km
54.000 in the State of Telangana under NHDP Phase –III on Design, Build, Finance, Operate
and Transfer (DBFOT) Toll Basis Package No. NHDP-III/BOT/AP-04 it is entrusted to NHAI
PUI Warangal and the project cost is Rs. 488 Cr. M/s. Hyderabad Yadgiri Tollway Pvt. Ltd
was selected as contractor for Design, Build, Finance, Operate and Transfer (DBFOT) for a
length of 38.6 km highway and for a concession period of 23 years including 650 days of
Construction period. They commenced the project on 30.07.2010 during the execution period
they claimed 72,29,92,708.00 Rs due to Compensation for Non-Handing Over Of Land and
payments due on account of utility shifting ,which falls in the category of Non responsive
owner and time phasing and requisite contractor legislation.
Therefore the 1st and 4th identified factors were the key factors responsible for delays and
claims for NH163.
5.2. Description of NH 65 Case Study
GMR infrastructure limited led consortium was selected for NH 65 Hyderabad to Vijayawada
highway project on built operate and transfer (Toll) basis through international competitive
bidding route, the project faced a lot of challenges due to heavy built up area but connected on
schedule by good time phasing, client contractor partnering, quality and lastly project
management skills.
Prediction of Construction Dispute Using Artificial Neural Networks Testimonies From Indian
Construction Projects
http://www.iaeme.com/IJCIET/index.asp 592 [email protected]
They started the project on 9th October 2009 with NHAI as client and the project cost of
1740rs for a length of 181.5km and for a concession period of 15 years-4 lanes and 25years-
for 6 lanes, Toll revenue is share with NHAI during the execution of project there no claims,
no delays were encountered and so the project could completed the within the budget and time.
Therefore NH65 was a successful highway project with no delays and claims. This could
be because of good time phasing, client contractor partnering, quality and lastly project
management skills. Not all BOT projects will have delays but the key factors identified from
the study were found responsible for delays and claims for NH44 and NH 163.
6. SPEARMAN’S CORRELATION ANALYSIS
Spearman’s correlation analysis was used to identify the factors that have a high degree of
correlation with the dispute attribute using Statistical package for social sciences (SPSS 23.0).
Correlation analysis was preferred as correlation coefficient measures the strength of
association between a pair of random variables. Based on the results of the spearman’s
correlation analysis (Table VI) the factors such as Time phasing and requisite contracting
legislation, Project financials and client contractor partnering, Quality and risk management
under ambiguity, Non responsive owner and unrealistic contractor rules were found to be
significantly correlated.
Table 7. Spearmen correlation for Project dispute attributes
Pd
a8
Pd
a4
0
Pd
a3
4
Pd
a4
Pd
a4
5
Pd
a4
6
pd
a5
0
pd
a5
1
pd
a2
0
pd
a4
8
pd
a4
7
pd
a1
9
pd
a1
4
pd
a1
5
pd
a1
0
pd
a2
9
pd
a2
7
pd
a2
3
O
Pd
a8
1
Pd
a4
0
.25
2
1
Pd
a3
4
-.01
8
-.17
3
1
Pd
a4
-.02
2
-.16
5
.62
4**
1
Pd
a4
5
-.01
4
.85
4**
-.30
5
-.35
6
1
Pd
a4
6
.15
5
.78
4**
-.38
3*
-.40
3*
.84
2**
1
pd
a5
0
.14
0
.52
8**
.33
8
.46
7**
.30
3
.10
2
1
pd
a5
1
.21
0
-.19
7
.52
8**
.61
2**
-.39
0*
-.42
7*
.34
5
1
pd
a2
0
.02
1
-.23
8
.42
2*
.75
5**
-.43
6*
-.38
1*
.33
3
.49
9**
1
pd
a4
8
-.09
6
-.09
6
-.16
6
-.41
1*
.00
5
.20
5
-.38
4*
-.29
9
-.29
4
1
pd
a4
7
-.09
2
.57
9**
-.56
1**
-.60
3**
.67
4**
.71
6**
-.24
2
-.67
8**
-.63
1**
.21
7
1
Asra Fatima, Dr. Bellam Sivarama Krishna Prasad and Dr. T.Seshadri Sekhar
http://www.iaeme.com/IJCIET/index.asp 593 [email protected]
pd
a1
9
-.22
4
-.66
9**
.45
3*
.36
2*
-.61
1**
-.58
1**
-.24
9
.44
8*
.25
0
.27
9
-.50
6**
1
pd
a1
4
-.02
7
-.37
2*
.49
1**
.49
4**
-.48
6**
-.68
6**
.31
1
.65
1**
.47
2**
-.35
2
-.65
1**
.33
0
1
pd
a1
5
-.31
7
-.80
4**
.13
8
.11
4
-.62
5**
-.52
3**
-.47
4**
.05
6
.11
5
.35
9
-.35
2
.83
9**
.13
5
1
pd
a1
0
.04
5
.68
6**
-.41
8*
-.45
8*
.76
8**
.82
6**
.11
1
-.58
8**
-.40
1*
.19
1
.77
9**
-.56
8**
-.54
6**
-.36
2*
1
pd
a2
9
-.13
3
-.71
3**
.22
9
.16
8
-.65
2**
-.53
3**
-.50
0**
.22
4
.11
0
.21
0
-.29
4
.80
2**
.17
7
.84
2**
-.38
4*
1
pd
a2
7
.23
9
.04
2
.13
2
-.10
3
.07
6
.03
8
-.33
4
-.04
3
-.17
5
.20
9
.11
0
.05
2
-.07
6
-.11
6
-.16
3
.01
8
1
pd
a2
3
.15
3
-.09
2
.47
6**
.69
3**
-.38
1*
-.29
4
.43
9*
.29
8
.56
4**
-.27
5
-.48
0**
.05
0
.32
8
-.02
4
-.25
1
.00
8
-.20
3
1
O
.21
6
.90
**
-.23
1
-.17
3
.84
4**
.76
9**
.53
1**
-.27
6
-.17
8
-.17
6
.56
4**
-.80
3**
-.34
9
-.85
4**
.70
7**
-.80
2**
-.00
9
-.04
2
1
7. SUMMARYAND CONCLUSIONS
The purpose of this study was to identify the critical factors that affect the project dispute
criterion which could be used to develop a project dispute prediction model. The conclusions
derived from the study are given below: Models to predict the project dispute criterion have
been developed through correlation studies (see Table IV) and with the usage of ANN in
MATLAB. When project dispute attributes is the prime objective, the following are the factors
that have been found significant Time phasing and requisite contracting legislation, Project
financials and client contractor partnering, Quality and risk management under ambiguity, Non
responsive owner and unrealistic contractor rules. This is consistent with the findings of Ling
et al. (2007) and Tam and Le (2007) that adequate training and good coordination are a must
for enhanced performance. The developed ANN models had a feed forward network based on
back propagation algorithm and the 5-9-1 structure gave the least MAPD of 12.22percent.The
high degree of predictive ability shows that the factors identified from correlation analysis are
correct and can be used to predict the project dispute attributes.
The project professionals can concentrate on certain factors instead of handling all the
factors at the same time to achieve the desired objectives. The study may be helpful to the
project manager and his team to predict the project dispute attributes during its course. Further
study In order to provide simple access to the developed prediction model, an interface may be
developed to promote data input and also to know the predictions. As such the user interface is
being developed and once completed it would be able to help project team members to feed the
factor inputs directly so as to know the corresponding project dispute attributes.
Prediction of Construction Dispute Using Artificial Neural Networks Testimonies From Indian
Construction Projects
http://www.iaeme.com/IJCIET/index.asp 594 [email protected]
REFRENCES
[1] Anderson, S.D. (1992), ‘‘Project quality and project managers’’, International Journal of
Project Management, pp.138-44.
[2] Goh, A.T.C.(1995), ‘‘Back propagation neural networks for modeling complex systems’’,
Artificial Intelligence in Engineering, Vol.9, pp.143-51.
[3] Charles T. Jahren, Member, ASCE, and Bruce F. Dammeier” Investigation into
construction disputes”, J. Manage. Eng. 1990.6:39-46.
[4] Edwards, S.R. (2007), ‘‘Modelling perceptions of building quality – a neural network
approach’’, Building and Environment, Vol. 42,pp.2762-77
[5] Ling, F.Y.Y. and Liu, M. (2004a), ‘‘Using neural network to predict performance of design-
build projects in Singapore’’, Journal of Building and Environment,Vol. 39, pp.1263-74.
[6] S. O. Cheung, C. M. Tam, and F. C. Harris (2000)” Project dispute resolution
satisfaction classification through neural network.” J. Manage. Eng., ASCE.,16,70-
79.
[7] McCulloch and Pitts (1943), “Fuzzy Logic and Neural Network Handbook”,
McGraw-Hill, Inc., New York, USA, 221-235.
[8] Ibrahim Yitmen & EbrahimSoujeri (2010) “An artificial neural network model for
estimating the influence of change orders on project performance and dispute
resolution”.,Proceedings of the International Conference onComputing in Civil and
Building Engineering.
[9] K.N. Jha and C.T. Chockalingam(2009)” Prediction of quality
performanceusingartificial neural networks Evidence from Indian construction
projects” Journal of Advances in Management Research Vol. 6 No. 1, 2009 pp. 70-
86
[10] Jain, A.K., Mao, J. and Mohiuddin, K.M. (1996), ‘‘Artificial neural networks: a
tutorial’’, IEEE (theme feature), Vol. 29 No. 3, March,pp.31-44
[11] Zurada, J.M. (1992), Introduction to Artificial Neural Systems,West Publishing,Los
Angeles, CA.
[12] N. B. ChaphalkarandSayali S. Sandbhor(2015)”Application of Neural Networks in
Resolution of Disputes for Escalation Clause Using Neuro-Solutions,”KSCE
Journal of Civil Engineering (2015) 19(1):10-16