An Improved Global Particle Swarm Optimization for Faster Optimization Process

Abstract

An efficient Global Particle Swarm Optimization (GPSO) is proposed in order to overcome the concern of trapping in the local optimal point especially in high dimensional while using ordinary Particle Swarm Optimization (PSO). GPSO is able to bring all the particles to be closely clumped together faster than PSO. In this paper, an improved GPSO is proposed in order to get a closely clumped particles group faster than using GPSO. The original GPSO is improved by taking into account the global best fitness error and particle fitness clumping size of every iteration. The improved GPSO is simulated by using several two dimension mathematical function and benchmarked with the original GPSO. The improved GPSO is shown to be able to obtain closely clumped particles much more faster than the original GPSO up to 62%. The performances are also evaluated by comparing the standard deviation, average, best particle and worst particles obtained through a 50 independent runs. In term of the four factors mentioned, the improved GPSO performance is shown to be as good of the original GPSO.