Research On Text Clustering Based On Dual Particle Swarm Optimization And K-means

K-means clustering algorithm is an important method in clustering field, but so strong dependence on the initial cluster centers of it that the clustering results may not be ideal. The characteristics of PSO and k-means algorithm are analyzed in this thesis, and particle swarm optimization was selected as a basis algorithm to lower sensitivity of k-means algorithm on the initial cluster centers. Established the basic framework for the evolution of dual population, and designed self-regulating mechanism of inertia weight which is based on the evolution of the population automatically adjust the degree of inertia weight. The exchange mechanism information was designed and a dual particle swarm optimization was proposed. An algorithm is proposed by combing the dual particle swarm optimization algorithm with k-means algorithm and applied to text clustering. Two kinds of individuals encoding and corresponding population initialization methods were designed. Reciprocal of sum of the within-class scatter was the fitness function. Experimental results show that the dual particle swarm optimization algorithm has higher convergence rate and solution accuracy. The clustering result of algorithm based on the dual particle swarm optimization algorithm and k-means also has high accuracy, and it is an efficient and accurate clustering algorithm.