Clustering based fuzzy particle swarm optimization

In local version (lbest) of particle swarm optimization (PSO), each particle has only the information of its own and its neighbors´ best, rather than all the population. The neighborhood of each particle is generally defined as topologically nearest particles to such particle at each side. There is no robust approach for determining the neighborhood size in the literature and all of them rely on try and error. In this paper, we introduce a new approach for defining neighborhood and propose a clustering based fuzzy particle swarm optimization (CFPSO). Our model is capable of finding the optimum number of neighborhood and also allows several particles to effect each other. We test our model on three test functions and compare the results with the global version of PSO (gbest) and two different topologies of lbest.