An Interval-Based Approach for Evolutionary Multi-Objective Optimization of Project Portfolios

Abstract

Project portfolio selection is addressed here as a multi-objective optimization problem. This work introduces an interval-based method that takes into consideration imperfect knowledge of the contribution of projects to a portfolio, the project requirements, available resources and preference parameters in the model. The multi-objective optimization problem is solved using an evolutionary algorithm that is adapted to handle intervals. To direct the search toward the region of interest of the Pareto frontier, the preferences of the decision maker (DM) are incorporated using an interval-based outranking approach. This allows to address problems with many objective functions; intransitive preferences and incomparability situations can also be handled using this approach. In terms of analyzing robustness, the DM can obtain different solutions according to his/her level of conservatism. The effectiveness of this proposal was tested both on an example from the related literature and another example of a public project portfolio with nine objective functions and large number of applicant projects.

An Interval-Based Evolutionary Approach to Portfolio Optimization of New Product Development Projects

“…Setting precise values of the conservatism-related thresholds is certainly demanding for the DM. Parameters and were introduced by Balderas et al [26] in the frame of interval-based project portfolio optimization. The present paper has introduced to model the DM' attitude facing the risk related to time effects.…”

“…The present paper has introduced to model the DM' attitude facing the risk related to time effects. As stated by Balderas et al [26], those conservatism parameters should be set in a coconstructive process involving a DM-decision analyst pair. In terms of the conservatism thresholds, an interaction between the DM and the decision analyst is mandatory in which the DM should understand the meaning of -feasibility and -dominance.…”

“…In terms of the conservatism thresholds, an interaction between the DM and the decision analyst is mandatory in which the DM should understand the meaning of -feasibility and -dominance. In [26] the reader can find illustrative examples of this interaction. The concept of a -feasible portfolio is easier than non--dominance; hence, the value of should be set first; this setting can be achieved by comparing the interval number associated with the available resources with different interval numbers representing potential levels of resource consumption (see [26]).…”

“…The uncertainty of projects is related to the degree of precision with which the variations in outcome, resources, and work processes of projects can be forecast [21]. From the point of view of multicriteria decisions, other authors prefer to use the more general term "imperfect knowledge" (e.g., [22][23][24][25][26]); the sources of this imperfect knowledge are arbitrariness, imprecision, illdetermination, and uncertainty in data and model parameters [22]. In this sense, the term "uncertainty" is related to the values of certain data in a more or less distant future [22].…”

“…[30][31][32][33][34][35]). Interval analysis combined with outranking methods was recently applied to portfolio optimization in [24,26]. The interval approach is a natural and simple way in which to express imperfect information, it does not matter its source.…”