摘要
现有数据流聚类算法只能处理Ti me Series和Cash Register型数据流,并且应用于高维数据流时其精度不甚理想。提出针对高维Turnstile型数据流的子空间聚类算法HT-Stream,算法对数据空间进行网格划分,在线动态维护网格单元信息,采用倾斜时间窗口存储统计信息,根据用户指定时间跨度离线输出聚类结果。基于真实数据集与仿真数据集的实验表明,算法具有良好的适用性和有效性。
Previous method only can deal with Time Series and Cash Register data stream. Moreover, the efficiency of clustering high dimensional data stream is not very satisfactory. In this paper a novel algorithm for clustering Turnstile data stream named HT-Stream is presented. HT-Stream partitions the space into grids, summarizes statistical information over data stream according to the tilted time window, and finds the clusters offline. HT-Stream can resolve high dimensional clustering problem and discover clusters with arbitrary shape. The experimental results on real datasets and synthetic datasets demonstrate promising availabilities of the approach.
出处
《计算机科学》
CSCD
北大核心
2006年第11期14-17,37,共5页
Computer Science
基金
国家自然科学基金(70371015)
教育部高等学校博士学科点科研基金(20040286009)
江苏省高校自然科学计划一般项目(05KJB520022)资助
关键词
数据流
子空间聚类
高维
倾斜时间窗口
Data stream,Subspace clustering, High dimension, Tilted time windows