Multicriteria Decision Analysis Approach for Stochastic Ranking with Application to Network Resilience

Abstract

Stochastic ranking, or the ordered comparison of alternatives whose performance is measured with random variables, is often made by relating expected values. Doing so ignores variability in performance, leaving the possibility of choosing an alternative that performs well on average but perhaps poorly in a worst-case scenario. Despite the several approaches and metrics that have been used to look beyond the expected value when comparing random variables, the possibility of emphasizing specific distribution percentiles has received little attention. Thus, a hybrid multicriteria decision analysis technique for the stochastic ranking of alternatives is proposed, giving consideration to selected performance points of the alternatives (e.g., typical and worst case). The approach, based on Hasse diagrams and the preference ranking organization method for enrichment evaluations (PROMETHEE), is illustrated with an example in network resilience, where link restoration times are stochastic. Restoration strategy rankings differ when typical and worst-case restoration times are explicitly considered relative to the standard expected value.