一种基于多重聚类的离群点检测算法  

Multi-clustering based outlier detect algorithm

作　　者：古平[1] 刘海波[1] 罗志恒[1] 

机构地区：[1]重庆大学计算机学院,重庆400044

出　　处：《计算机应用研究》2013年第3期751-753,756,共4页Application Research of Computers

基　　金：中央高校基本科研业务费科研专项基金资助项目(CDJZR 10180008)

摘　　要：在LDOF算法的基础上,提出一种基于多重聚类的离群点检测算法PMLDOF。该算法针对局部离群度量计算量大的缺点,采用聚类剪枝技术作为减少计算量的方法;同时,为了避免将位于簇边缘的离群点错剪,算法利用多重聚类的差异性对簇的边缘点进行筛选。在对数据集进行剪枝后,计算剩余数据的局部离群度LDOF,并找出符合条件的离群数据点。实验结果表明,算法在时间复杂度和检测精度上具有更好的优越性。Based on the LDOF algorithm, this paper proposed a multi-clustering based outlier detection algorithm PMLDOF. To reduce the amount of calculation for local outlier factor, it employed cluster pruning technique. The other improvement was to filter the non-outliers based on the difference of multiple clustering, which was to avoid the error pruning of outliers located at the edge of clusters. After pruning, it calculated the local outlier factors of remaining data and obtained the outliers. Experi- mental results show PMLDOF is superior to LDOF algorithm not only in the efficiency but also in the detection accuracy.

关 键 词：数据挖掘 离群检测 剪枝 多重聚类 局部离群度 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]