Robust Portfolio Optimization in Multi-Criteria Project Selection

Selection of a project portfolio with limited resources is an important decision in both public and industrial organizations. Often these decisions are complicated by a high number of project proposals, considerations of multiple criteria and resources, project interactions, uncertain project outcomes and incomplete information about decision maker's (DM's) preferences. Moreover, opportunity costs of funding projects often make choice of the total resource expenditure an integral part of project portfolio selection. This thesis presents a decision analytic framework, entitled Robust Portfolio Modeling (RPM), whose novelty lies in i) the modeling of incomplete information about the DM's preferences and projects' properties and ii) identifying robust project and portfolio decisions in view of such incomplete information subject to varying levels of resource expenditure. Under such incomplete information, the concept of a single optimal portfolio is generalized to the set of non-dominated portfolios. Based on the computation of nondominated portfolios, robust core projects included in all non-dominated portfolios can be identi ed. Similarly, exterior projects are not included in any non-dominated portfolios and can thus be discarded from further analysis. Thus, e orts of obtaining additional information can be focused on the remaining borderline projects to reduce the number of non-dominated portfolios and to identify more core/exterior projects. The nal portfolio can be structured by selecting borderline projects based on holistic evaluation, through negotiation in group decision settings or by relying on decision recommendations that identify robust portfolios which perform well even in the worst case realization of the model parameters. The use of RPM is illustrated through a case study in product portfolio selection. Number of pages: 92