Abstract: The traditional clustering algorithms have many shortcomings, such as sensitive to initial value and vulner-
able to local minima. The method to determine the number and location of cluster center is proposed. Cloud theory is used to transform particle swarm optimization to improve the performance of PSO. This methed can search more reasonable clustering center. Experimental results show that the algorithm has solved these two drawbacks and obtained a good stability and a better clustering result.

Key words: clustering algorithm, particle swarm, fuzzy c-means

摘要： 针对传统的聚类算法存在对初始化值敏感和容易陷入局部极值等缺点，提出一种确定聚类中心数目和位置的方法。用每一个粒子表示一组聚类中心，采用云理论改造粒子群算法，从而提高粒子群算法的性能，以便搜索到更合理的聚类中心完成聚类划分。实验结果表明该算法很好克服了这两个缺点，获得了稳定性好和更紧凑的聚类效果。

关键词: 聚类算法, 粒子群, 模糊C均值