Clustering-based selection for evolutionary multi-objective optimization

Abstract

In this study, a novel clustering-based selection strategy of nondominated individuals for evolutionary multi-objective optimization is proposed. The new strategy partitions the nondominated individuals in current Pareto front adaptively into desired clusters. Then one representative individual will be selected in each cluster for pruning nondominated individuals. In order to evaluate the validity of the new strategy, we apply it into one state of the art multi-objective evolutionary algorithm. The experimental results based on thirteen benchmark problems show that the new strategy improves the performance obviously in terms of breadth and uniformity of nondominated solutions.