摘要： 传统小波聚类算法标记满足密度阈值的连通单元为同一个簇,而不满足密度阈值的网格有可能存在属于簇的数据对象,数据的每维属性有时差距较大,不合适再划分均匀网格。为此,提出一种改进的小波聚类算法CWaveCluster,划分非均匀网格,进一步细化边界网格,对不满足密度阈值的网格进行处理,最终形成聚类。在指定的快速存取记录器(QAR)数据集上的实验结果表明,改进的小波聚类算法能根据数据特点划分网格,区分簇与簇的边界,有效解决QAR数据异常点检测问题。
关键词:
连通单元,
小波聚类,
边界网格,
快速存取记录器,
密度阈值
Abstract: Traditional wavelet clustering algorithm labels the communication unit satisfying density threshold as the same cluster,the mesh which does not meet the density threshold may have the data objects belonging to the cluster,and each dimension attribute of the data sometimes has a big gap,so that subdividing the mesh into uniform grid is not appropriate.Thus,an improved wavelet clustering algorithm is proposed.The method is used to divided the non-uniform grid,and refines further the boundary of the grid which does not satisfy the density threshold,and formats the clusters finally.By applying on the specified Quick Access Recorder(QAR) data sets,experimental results show that the improved wave cluster algorithm can effectiveing distinguish between cluster and boundary of the cluster,according to the characteristics of the data mesh,this method solves the question of the QAR data anomaly detection effectively.
Key words:
communication unit,
wavelet clustering,
border grid,
Quick Access Recorder(QAR),
density threshold
中图分类号: