Building Nearest Prototype Classifiers Using a Michigan Approach PSO

摘要: This paper presents an application of particle swarm optimization (PSO) to continuous classification problems, using a Michigan approach. In this work, PSO is used process training data find reduced set prototypes be classify the patterns, maintaining or increasing accuracy nearest neighbor classifiers. The approach represents each prototype by and uses modified movement rules with competition cooperation that ensure diversity. result particles are able recognize clusters, decision boundaries achieve stable situations also retain adaptation potential. proposed method tested both artificial problems three real benchmark quite promising results