一种领域覆盖的数据流聚类算法  

Data Stream Clustering Algorithm Based on Neighborhood Covering

作　　者：章季阳[1] 王伦文[1,2] 

机构地区：[1]电子工程学院309室,合肥230037 [2]通信信息控制和安全技术重点实验室,浙江嘉兴314033

出　　处：《小型微型计算机系统》2012年第9期1913-1916,共4页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(60675031)资助

摘　　要：数据流聚类分析是数据流挖掘的重要手段之一.为满足数据流不断演化及高速处理的要求,提出一种领域覆盖的数据流聚类算法NCStream(Stream clustering algorithm based on Neighborhood Covering).该算法通过建立领域覆盖模型,详细定义和分析了数据流演化过程中覆盖簇调整、创建、删除和合并的行为操作,并同时对覆盖簇的聚类特征予以在线维护.与同类算法相比,NCStream算法无需事先指定聚类簇数,避免参数设置对聚类结果造成的影响,而且易于建立空间索引,因此能够更加有效地反映数据流的演化情况.实验采用无线电实际监测数据集构造数据流,实验结果表明NCStream算法在聚类形状、聚类质量以及处理时间方面具有更好的性能.Data stream clustering analysis is one of the key techniques in data stream mining. To meet the requirement of evolution and high- speed processing, a data stream clustering algorithm based on Neighborhood Covering is proposed, namely NCStream. By building Neighborhood Covering model, the proposed algorithm for the evolving procedure of data stream is defined and analyzed at length, including the adjustment, creation, deletion and mergence of covering cluster, and simultaneously maintain the cluster feature online. Compared with the similar clustering method, NCStream has no assignment in the number of cluster in advance, which avoids the disadvantage of clustering result due to parameter setting. Moreover, NCStream benefits the establishment of spatial index. Hence, the evolution of data stream is more effectively reflected. The experimental results on real wireless monitor data sets demon- strate that NCStream is of better performance in clustering shape, quality and processing time.

关 键 词：数据流 聚类 领域覆盖 演化 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]