Bayesian-based indifference-zone multi-objective ranking and selection procedures

Abstract

Three indifference-zone (IZ) multi-objective ranking and selection (MORS) procedures are presented in this paper. The procedures define the minimum number of simulation replications required for each alternative system which will guarantee that the probability of correct selection exceeds a desired minimum. The statistical validity of the procedures are shown by using a Bayesian inference model, which contributes to the efficiency of the procedures. The proposed procedures were compared to multi-objective budget allocation procedures and performed well in numerical experiments. The concept of relaxed Pareto optimality is also proposed, which incorporates the indifference-zone concept into the multi-objective optimization domain.