论文部分内容阅读
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CWave Cluster,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。
The traditional wavelet clustering algorithm marks the connected cells that meet the density threshold as the same cluster, while the grids that do not meet the density threshold may have data objects belonging to the cluster. grid. To this end, an improved wavelet clustering algorithm, CWave Cluster, is proposed to divide non-uniform grids, further refine the boundary grids, process the grids that do not meet the density thresholds, and finally form a cluster. The experimental results on the specified QAR datasets show that the improved wavelet clustering algorithm can divide the grids according to the characteristics of the data and distinguish the boundaries of the clusters from the clusters, so as to effectively solve the problem of detecting the outlier of QAR data.