Promethee-i Selecting the Best Simulation Model Configuration Based on Multiple Performance Measures

We introduce the use of a variant of the original multicriteria decision making method Promethee, in order to select the best simulation model configuration among a finite set of alternatives. For each alternative configuration a number of replications of terminating simulation runs is performed. At the end of each replication, and for each configuration, the result of a number of performance measures is obtained. In the selection problem, these performance measures are typically conflicting criteria for which the alternative configurations have been assessed by a number of computer simulation replications. We submit these data to an interval version of the Promethee outranking method, in order to select the best model configuration. We illustrate this by means of an incident management model for a call centre.