分布式密度和中心点数据流聚类算法的研究

摘要

分析分布式数据流聚类算法的基本框架结构，针对CluStream算法对非球形聚类效果不佳提出一种基于密度和中心点的分布式数据流聚类算法DDCS-Clustering（Distributed Density and Centers Stream Clustering）。该算法应用密度、中心点与衰减时间窗口，在分布式环境下对数据流进行聚类。实验结果表明，DDCS-Clustering算法具有较高的聚类质量与较低的通信代价。%We analyse the basic structure of distributed data stream clustering algorithm,and propose a kind of distributed data stream clustering algorithm which is based on density and centre points named the DDCS-clustering aiming at that the CluStream algorithm is less effective for non-spherical cluster.The algorithm applies the density,centre points and decay time windows mechanism,clusters the data stream under the distributed environment.Experimental result shows that the DDCS-clustering algorithm has higher clustering quality and lower communication cost.