摘要
提出一种基于滑动窗口的概率数据流聚类方法PWStream。PWStream采用聚类特征指数直方图保存最近数据元组的信息摘要,在允许的误差范围内删除过期的数据元组;并针对数据流上概率元组提出强簇、过渡簇和弱簇的概念,设计了一种基于距离和存在概率的簇选择策略,从而可以发现更多的强簇。理论分析和实验结果表明,该方法具有良好的聚类质量和较快的数据处理能力。
An effective clustering algorithm called PWStream for probabilistic data stream over sliding window is developed.The algorithm uses exponential histogram of cluster feature to store the summary information of the most recently arrived tuples,and outdated information is deleted within a certain guaranteed range of error.For the uncertain tuples in data stream,the concepts of strong cluster,transitional cluster and weak cluster are proposed in the PWStream.With these concepts,an effective strategy of choosing cluster based on distance and existence probability is designed,which can find more strong clusters.Theoretical analysis and comprehensive experimental results demonstrate that the proposed method is of high quality and fast processing rate.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第4期141-145,共5页
Computer Engineering and Applications
基金
安徽省自然科学基金(No.090416247,No.070412055)
安徽省高校自然科学研究计划项目(No.KJ2009B139)
安徽省高等学校青年教师科研资助计划项目(No.2008jq1143)~~
关键词
概率数据流
聚类
滑动窗口
直方图
probabilistic data stream
clustering
sliding window
histogram
