Differential evolution with nearest better clustering for multimodal multiobjective optimization

Abstract

The main aim of multimodal multiobjective optimization algorithms is to find multiple optimal solutions for multiple objectives. The problem with multiple objectives is to maintain the diversity and convergence balance in both decision and objective space. Better performance in one space can lead to deterioration of performance in another space. To effectively perform exploration and exploitation in the decision space, we have proposed a niching technique in our proposed algorithm based on differential evolution to form groups of the population. The niching technique also incorporates a balancing strategy to average the size of niches and a generation strategy to add new elements by finding promising areas. We have also proposed an archive strategy to store the possible optimal solutions that help to explore the decision space and ensure diversity as well as the convergence of the solutions. To show the efficiency of our proposed algorithm, we have performed exhaustive experimentation on the benchmark test suite and compared the results with state-of-the-art algorithms. The experimental results obtained using the proposed algorithm highlight the efficiency and effectiveness of our proposed algorithm in both decision and objective space.