Real-World Parameter Tuning using Factorial Design with Parameter Decomposition

Aldy GUNAWAN, Singapore Management University
Hoong Chuin LAU, Singapore Management University
Elaine WONG, Singapore Management University

Abstract

In this paper, we explore the idea of improving the efficiency of factorial design for parameter setting of metaheuristics. In a standard full factorial design, the number of runs increases exponentially as the number of parameters. To reduce the parameter search space, we propose a decomposition of factorial design where parameters are partitioned into disjoint categories. One approach is to classify them as important and less-important parameters based on user guidance. Another proposed approach is to apply a fractional factorial design to partition parameters based on their main effect values. We empirically evaluate our approach against an existing automated parameter tuning configurator, namely ParamILS, to tune a simulated annealing algorithm for a real world spares inventory optimization problem. We conclude that our proposed methodology leads to improvements in terms of the quality of the solutions compared against the default para-meter values.