A Meta Model Based Particle Swarm Optimization for Enhanced Global Search

Abstract

The performance of a metaheuristic algorithm depends on the appropriate selection of its behavioral parameters. A good selection of parameters increases the search ability of an algorithm and avoids premature convergence. Particle swarm optimization (PSO) is swarm-based metaheuristic algorithm, which needs few parameter adjustments and less computational time. Meta-optimization has been used to tune the parameters and to get better results. Previously, authors applied meta optimization techniques to specific problems to tune the parameters and to get better results for specific case studies in different fields, but the application of meta optimization in benchmark functions are limited. The present study proposes meta optimization-based PSO to minimize the computational effort required for manual trial and error-based parameter selection. The proposed algorithm is tested for 14 benchmark functions (8 unimodal and 6 multimodal), and its efficiency and robustness are assessed via statistical analysis. The algorithm outperforms other renowned established algorithms (GA, PSO), and its performance remains consistent with increasing modality and dimensionality.