Abstract: Because the K-means clustering method is sensitive to the selection of the initial clustering center and it may converge to local optima. A new particle swarm optimization clustering method is proposed. The method introduces the improved crossover and mutation operations of the genetic algorithm, which can maintain the diversity and quality of varieties of the particles, reduce the impact of the random initial cluster centers. Furthermore, it combines with particle swarm optimization increasing the particle swarm’s global search capability. Experimental results show that the proposed method improves the stability and the accuracy of classification.

Key words: crossover and mutation operator, Particle Swarm Optimization（PSO）, cluster, K-means

摘要： 针对K-均值聚类方法受初始聚类中心影响，容易陷入局部最优解的问题。提出了一种新的粒子群优化聚类方法，该聚类方法采用改进的交叉、变异算子，使群体粒子保持品种的多样性和优良性，减小随机初始聚类中心的影响，同时结合粒子群优化算法，增加粒子群的全局搜索能力。实验结果表明，提出的方法在稳定性和分类准确率上都有所提高。

关键词: 交叉变异算子, 粒子群优化（PSO）, 聚类, K-均值