Pareto set estimation with guaranteed probability of correct selection

We consider the ranking and selection problem with multiple objectives and present a procedure to estimate a Pareto set. We provide three different formulations, namely, simultaneously retaining all desirable systems and eliminating all undesirable systems, or achieving either one of the two aforementioned goals. We address situations where all systems and objectives have either independent or correlated observations (e.g., due to the use of common random numbers). In each case, we identify appropriate choices of parameter values and prove that the resulting algorithm guarantees the desired probability of correct selection in a finite amount of time. Numerical experiments are provided to support the validity and efficiency of the algorithms. (C) 2020 Elsevier B.V. All rights reserved.

Automated summary

This study addresses the ranking and selection problem with multiple objectives by proposing three different formulations and verifying the effectiveness and efficiency of the algorithms through numerical experiments. The appropriate choices of parameter values and the guarantee of correct selection probability in a finite amount of time are demonstrated in each scenario.