基于MPI的并行PSO混合K均值聚类算法

摘要

The performance of traditional serial clustering algorithm cannot meet the needs of data clustering of the huge amounts of data.To enhance the performance of clustering algorithm, a new clustering algorithm combining parallel Particle Swarm Optimization (PS0) with K- means based on MPI was proposed in this paper.Firstly, the improved PS0 was combined with K- means to enhance the capacity of global search, and then a new parallel clustering algorithm was proposed, which was compared with K- means and PSO clustering algorithms.The experimental results show that the new algorithm has better global convergence, and also has higher speed-up ratio.%传统的串行聚类算法在对海量数据进行聚类时性能往往不尽如人意,为了适应海量数据聚类分析的性能要求,针对传统聚类算法的不足,提出一种基于消息传递接口(MPI)集群的并行PSO混合K均值聚类算法.首先将改进的粒子群与K均值结合,提高该算法的全局搜索能力,然后利用该算法提出一种新的并行聚类策略,并将该算法与K均值聚类算法、粒子群优化(PSO)聚类算法进行比较.实验结果表明,该算法不仅具有较好的全局收敛性,而且具有较高的加速比.