POLIBITS, Vol. 62, pp. 77-84, 2020.
Abstract: Project Portfolio Selection (PPS) is a major strategic decision problem faced by any organization. PPS decides how to invest resources into projects subject to a decision process influenced by multiple conflicting criteria. The portfolio’s compromise to the organization’s well-being has an uncertainty that directly affects a decision maker’s preferences (DM). MOEA/D is a well-known approach to tackle multicriteria optimization problems, and it is still open for the development of strategies to handle uncertainty on its search process. This work proposes I-MOEA/D, a new method based on a MOEA/D approach, to deal with DM’s uncertainty in costs and benefits of portfolios’ projects. The proposed novel features include (a) handling large numbers of objectives; (b) a method to generate the initial population; and (c) handling the uncertainty of resources, costs, and benefits through intervals. An experiment compared I-MOEA/D against the state-of-the-art I-NSGA-II algorithm in instances with two to fifteen objectives. Results demonstrate the competitiveness of I-MOEA/D by improving the quality of solution of I-NSGA-II in most instances.
Keywords: Decision making, uncertainty, multi-objective optimization, mathematics of intervals, project portfolio problem
PDF: Optimization of Many Objectives with Intervals Applying the MOEA/D Algorithm
PDF: Optimization of Many Objectives with Intervals Applying the MOEA/D Algorithm
https://doi.org/10.17562/PB-62-9
See table of contents of POLIBITS 62.