Abstract
Particle Swarm Optimization (PSO) is one of the population-based metaheuristics. The algorithm is simple yet the convergence is rather fast. The parameter values have a large influence on the search performance. Many studies on PSO parameter tuning have been conducted so far. This paper proposes determining parameters separately in each stage into which the iterative process is divided, using Design of Experiments (DoE) methodology. The situation of the swarm is considered to be changing as the process proceeds. Hence, a certain effect is expected with this method. The exploration with the proposed method is performed on four test functions. Additionally the exploration with three other methods is also conducted for comparison. The proposed method yields good results. This study also reveals that the search performance improves when parameters, such as inertia weight, increase as the process proceeds.
