A Bayesian Network-based Decision Framework for Selecting Project Delivery Methods in Highway Construction

Abstract

Transportation agencies currently have several options in delivering their highway construction projects. Selecting an appropriate project delivery method (PDM) is a complex decision-making process. Researchers and transportation industry practitioners have been striving to discover the knowledge and methodologies to enhance the project delivery decision. However, through conducting an extensive literature review of existing methodologies, it is found that quantitative approaches, implementing probabilistic comparisons, to project delivery decisions are not fully addressed or understood. To fill this gap, this research aims at developing a decision framework by implementing Bayesian Network (BN), an advanced statistical tool, for selecting an appropriate PDM in highway construction industry. The BN-based decision framework incorporates the decision driving factors such as project attributes, risk profiles, project complexity, cost, and time. In developing the BN-based decision framework, this dissertation employed several research methodologies and techniques, including content analysis, questionnaire, case studies, cluster analysis, ANOVA, correlation and reliability analysis, and cross-validation techniques. The dissertation follows a four-journal paper format. The first paper explores the impact of project size on highway design-bid-build (D-B-B) and design-build (D-B) projects. The second paper identifies and evaluates the risks involved in highway project delivery methods: D-B-B, D-B, and construction manager/general contractor (CM/GC). Building upon the findings and results from the first two papers, the third paper determines the probabilistic dependence between the decision factors and develops a theoretical decision framework using BNs for selecting an appropriate PDM. The fourth paper focuses on demonstrating the practical application of the proposed BN-based decision framework using case studies. Also, the final paper presents a k-fold (cross-validation) technique to test and verify the accuracy of the proposed BN-based decision framework. This dissertation contributes to the theoretical body of knowledge by introducing a new quantitative approach using BNs for PDM selection. The findings from this study indicate that implementing BNs facilitate the owner/decision maker in a better understanding of probabilistic comparison and selection of an appropriate PDM for highway construction projects. State transportation agency officials can utilize these findings as a supplemental tool for their project delivery decisions.