With increasing demand for a more efficient transportation system and decreasing budget levels, transportation investment decision-making that aims to select the optimal project portfolio which... Show moreWith increasing demand for a more efficient transportation system and decreasing budget levels, transportation investment decision-making that aims to select the optimal project portfolio which yields maximized overall networkwide benefits in terms of economy, society and environment has increasingly become important. This dissertation has conducted an in-depth investigation into project evaluation and project selection that are crucial steps of transportation decision-making. It begins with information search through a review of existing methods for project evaluation and selection. Several limitations of existing methods have been revealed. In particular, they are in lack of considerations in network impacts of a single investment project, interdependencies of simultaneously implementing multiple projects, and restrictions of total risk of overall benefits of selected projects within an acceptable level. Then, a new methodology is proposed for networkwide traffic assignments, project evaluation, and project selection. A state-of-art large scale transportation simulation software, the TRansportation ANalysis and SIMulation System (TRANSIMS) toolbox, is utilized to perform networkwide dynamic traffic assignments to general redistributed traffic volumes after project implementation needed as inputs for project evaluation. For project evaluation, a life-cycle cost analysis approach is developed to consider all agency costs and user costs in the service life-cycle of two primary categories of highway facilities: pavements and bridges. In order to enhance the robustness of analytical results, risk and uncertainty of input factors concerning traffic volumes, project costs, and discount rates are incorporated into the life-cycle cost computation using @Risk Palisade software, Version 5.5. For project selection, two-stage enhanced Knapsack model, hypergraph Knapsack, and two-stage hypergraph Knapsack model are proposed to choose the best sub-collection of interdependent projects to yield maximized overall benefits at various budget levels, while controlling the total risk within an acceptable level. In terms of two-stage Knapsack model, the Markowitz mean-variance model is utilized for stage-one optimization to generate minimized total risk of all projects subject to constraints of available budget and minimum benefits to be expected for individual projects. At the second stage, the Knapsack model is enhanced by adding stage-one optimization solution as one more constraint. Such a treatment could help control the total risk of overall benefits of all selected projects at a desirable level. Moreover, a hypergraph Knapsack model is introduced to capture project network impacts and interdependency relationships. In order to simultaneously address issues of networkwide project impacts, interdependencies, and total risk levels, a two-stage hypergraph Knapsack model is developed. Efficient solution algorithms are developed and coded to Frontline Solver Xpress V55 software to solve the two-stage Knapsack model, hypergraph Knapsack model, and two-stage hypergraph Knapsack model, respectively. Three computational studies are performed to apply the proposed methodology using two sets of data, including six-year data on 672 candidate projects proposed by Indiana Department of Transportation for state highway programming and 6 mega projects proposed by Illinois State Toll Highway Authority for tollway network major capital improvements. It has generally found that the use of two-stage Knapsack model could readily control the total risk of overall benefits of selected projects at a desirable level, but it may result in significant changes in the overall benefits for different budget levels where significant differences in risks are associated with individual projects. The hypergraph Knapsack model could effectively handle issues of networkwide project impacts and interdependency relationships. However, the two-stage hypergraph Knapsack model appears to be most robust in that it could simultaneously resolve the issues of networkwide project impacts, interdependency relationships, and total risks of overall project benefits, thus generating most reliable information to support rational transportation investment decision-making. Show less