Parameter Optimization for Evolutionary Algorithm – Quadratic Assignment Problem

Abstract

Quadratic Assignment Problem (QAP) is a NP-hard problem. In order to eliminate binary decision variables and assignment constraint, the problem is formulated in a constraint logic programming in which the binary decision variables are now part of the constraint. Evolutionary algorithm is used to find the solution of the constraint logic programming. Mutation rate and population size are the parameters in the algorithm. Response Surface Technique is used to optimize those parameters. Response considered in this study is assignment cost (QAP objective value). Result shows that QAP in constraint logic programming substantially reduces computational intensity allowing the modest-size problems can be solved in Microsoft Excel 2010. Parameter optimization by means of response surface approach to locate optimal mutation rate and population size shows a promising trend. However, further study should be carried out to establish a solid ground for the parameter optimization methodology proposed in this study.