Abstract: K-means algorithm is simple and fast, however its result is affected by the initial clustering center and easily falls into the local optimum. This paper combines Particle Swarm Optimization(PSO) and adjusting mechanism and the immune memory function of immune system to improve K-means algorithm, and proposes a clustering algorithm based on Immune Particle Swarm Optimization algorithm(IM-PSO-KMEANS). The experiments show that the IM-PSO-KMEANS algorithm overcomes the problems of K-means algorithm, and the results of clustering are better than algorithm based on PSO.

Key words: clustering, Immune-PSO, K-means, Particle Swarm Optimization(PSO)

摘要： K均值算法简单快速，但其结果容易受初始聚类中心影响，并且容易陷入局部极值。该文结合粒子群优化算法和免疫系统中的免疫调节机制与免疫记忆功能对K均值算法进行改进，提出一种基于免疫粒子群优化的聚类算法。实验结果证明，该算法解决了K均值算法存在的对初值敏感的缺点，聚类结果稳定，而且比基于粒子群优化的聚类算法具有更好的聚类效果。

关键词: 聚类, 免疫粒子群优化, K均值, 粒子群优化

CLC Number: 

• TP311