An Dual-Clustering Search Strategy for Evolutionary Multi-Objective Optimization with Complicated Pareto Sets

In many real-world applications, decision makers tend to approach a good approximation to the Pareto optimal set (PS) for facilitating their decision making. This is still a challenge for most existing algorithms, particularly when solving multi-objective optimization problems with complicated PS (MOP-CPSs). To address this issue, a dual-clustering search strategy-based evolutionary algorithm has been tailored for handling MOP-CPSs. Two clustering-based operators are proposed to take full advantage of the inherent advantages of clustering techniques in mining the population data structure. Firstly, the local neighborhood relationship of the population is learned in the objective space to enhance the diversity maintenance. Then, these neighborhood relationships are used to construct a mating pool in the decision space to generate offspring, with the purpose of focusing on enhancing the exploitation ability of individuals. The proposed strategy can effectively promote the overall performance in both the decision space and objective space with population iteration. The experimental results on various MOP-CPSs demonstrate that our developed algorithm outperforms several state-of-the-art multi-objective evolutionary algorithms both competitively and effectively, and shows remark-able advantages in visualization.