Multicriteria decision support under uncertainty: combining outranking methods with Bayesian networks

Abstract Assessing alternative solutions that have uncertain evaluations in conflicting multiple criteria is not straightforward. Probabilistic models such as Bayesian networks (BNs) can effectively model and represent the uncertainty in such problems, but they do not include built-in mechanisms to guide different decision makers (DMs) with varying preferences toward the final decision. We propose a systematic approach to combine outranking methods with BNs to provide decision support for solutions with multiple and conflicting criteria under uncertainty. The proposed approach is based on Preference Ranking Organization Method for Enrichment Evaluation (PROMETHEE), and it offers different levels of precision and flexibility to the DMs in assessing the solutions. Our approach includes graphical tools and summary metrics to enhance the presentation of its results to the DMs. We test our approach with a case study on supplier selection where the uncertainty in supplier performances is modeled with a BN. We demonstrate that the proposed approach can enable the joint use of multiple criteria techniques with probabilistic modeling techniques like BNs to provide decision support in complex environments including uncertainty.