A Score Based Method for Controlling the Convergence Behavior of Particle Swarm Optimization

In recent years, Particle Swarm Optimization (PSO) has been used in data mining, feature extraction and other optimization based applications. Time to time, a number of researchers have suggested modifications to the basic PSO. Although this optimization technique finds good solutions much faster than the traditional and evolutionary algorithms, they suffer from a major drawback of premature convergence. In addition, it has been found experimentally that the quality of the solutions does not improve as the number of iterations is increased. In this paper we discuss the reason behind the premature convergence. We present a new method based on performance-scoring for improving the algorithm The scoring based model is applied to the basic and some of the modified versions of PSO models.