Abstract: Clustering can be regarded as the process of finding K optimal centers.A group of centers can be seen as a particle，and the inverse of the sum of scatter within class as optimal function，and then mutation probability is used as the condition of particle variation，so as to improve the ability of exploring and overcoming the shortcomings of the particle swarm converging to local optimization value.So the optimal cluster center can be found through mutation particle swarm optimization.The experiment shows that the clustering result of this algorithm is improved and it has good stability.

Key words: clustering, mutation, particle swarm optimization algorithm

摘要： 聚类可以看成是寻找K个最佳聚类中心的过程。把一组聚类中心视为一个粒子，把总类内离散度和的倒数看成优化函数，采用变异概率作为粒子变异的条件，从而提高了粒子群的探索能力，克服粒子群收敛到局部最优值的缺点。因此通过变异粒子群算法能够找到最佳聚类中心。实验结果表明该算法有很好的稳定性，提高了聚类效果。

关键词: 聚类, 变异, 粒子群算法