衰减窗口模型下基于密度的数据流聚类算法  

Density-Based Data Stream Clustering Based on Damped Window Model

作　　者：王冬秀[1] 

机构地区：[1]广西科技大学财经学院,广西柳州545006

出　　处：《科技通报》2013年第6期40-43,46,共5页Bulletin of Science and Technology

基　　金：广西教育厅科研项目(201106LX397)

摘　　要：对目前主流数据流算法的优缺点进行分析后,提出了一种衰减窗口模型下基于密度的数据流聚类算法DWDCluSteam。算法采用衰减窗口技术,然后利用改进的树结构来维护和更新数据流的摘要信息,最后利用周期性的剪枝策略,定期删除过期、稀疏的网格单元。仿真实验表明,相对于以往的数据流聚类算法,该算法可获得较好的聚类质量,较小的内存开销和较高的数据处理能力。To analysis the advantages and disadvantages of the recent main datastreams clustering algorithms,a densitybased data stream clustering over damped window model is put forward.DWDCluSteam uses damped window technology, then makes use of an improved tree structure to maintain and apdate the data stream summary,finally a period pruning strategy was disgned to delete expired and sparse grids.The simulation experiment suggests that the DWDCluSteam has better clustering quality and less memory consumption,higher data processing capacity.

关 键 词：数据流 聚类 衰减窗口 维度树 网格 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]