UPM Institutional Repository

Development of a hybrid method by data envelopment analysis and analytic network process for project selection


Citation

Mojahed, Majid (2014) Development of a hybrid method by data envelopment analysis and analytic network process for project selection. PhD thesis, Universiti Putra Malaysia.

Abstract

This research focuses on developing a model that can be used to rank construction projects and select the profitable one. This model will result from the integration of a decision tool called the Analytic Network Process and a data analysis model called Data Envelopment Analysis. Due to scarce resources, contractors cannot undertake all projects simultaneously. Because of the wide variety of criteria, they are always faced with difficulties in selecting projects, so a proper scientific method is necessary to aid contractors for achieving more profit. Data Envelopment Analysis (DEA) is a very powerful benchmarking technique that determines a Decision Making Units (DMUs) as efficient or inefficient units. Moreover, it can increase the performance of inefficient units by reducing the amount of input variables and increasing the output ones. A hybrid method of decision making has been developed using Analytic Network Process method (ANP) with DEA by considering the relationship between criteria and alternatives based on network system of elements for solving in Project Selection problems. The objectives of this research are identifying and determining the weights of significant criteria in order to rank construction projects based on network system of elements and also developing prototype software as a tool. In this study, two kinds of questionnaires have been applied. The first questionnaire was used to identify the relationship between elements and the second questionnaire specified the weight of 27 criteria. ‘Political impact in area’ was introduced as an important criterion for project selection. Both questionnaires were filled up by 26 contractors (whole group of contractors in elecommunication Company of NorthKhorasan) who are experts in 25 selected projects for Telecommunication Company. After classifying the criteria into corresponding groups based on their characteristics, the weight of each group was identified. The output of ANP method was modified to serve as input for DEA method. To identify project selection criteria and the number of groups, voting method was used. The inputs of Cross Efficiency Matrix were provided by inputs and outputs of DEA method and finally, this hybrid method represents the project number 3 ‘Building construction of call Centre in Jajarm’ was selected as the profitable project and it was followed by project number 5’ Building construction of call Centre in Farooj’ and project number 4‘Building construction of call Centre in Esfaraeen’ and so on. To show how much of this prioritization is close to reality, some ordinal numbers (rank of projects) were gathered based on each contractor’s viewpoint. Based on Cohen’s scales, the spearman’s rank correlation of 0.586 represents strong relationship between the results of the ranked projects by hybrid method and contractor’s viewpoints. The application of this hybrid method is not limited only to selecting and ranking projects. It is also applicable to any kind of decision making in the selection process. Prototype software is developed in the form of the MATLAB to help users in selecting and ranking projects. Even users without any knowledge of the applications like Super Decision and LINDO can solve these kinds of selection problems.


Download File

[img]
Preview
PDF
FK 2014 91RR.pdf

Download (1MB) | Preview

Additional Metadata

Item Type: Thesis (PhD)
Subject: Data envelopment analysis
Subject: Decision-making - Mathematical models
Call Number: FK 2014 91
Chairman Supervisor: Professor Rosnah Bt. Mohd. Yusuff, PhD
Divisions: Faculty of Engineering
Depositing User: Haridan Mohd Jais
Date Deposited: 19 May 2017 02:55
Last Modified: 19 May 2017 02:55
URI: http://psasir.upm.edu.my/id/eprint/50402
Statistic Details: View Download Statistic

Actions (login required)

View Item View Item