摘要
针对局部离群度量计算量大的缺点,在LDOF算法的基础上,提出一种新颖的基于聚类的离群点检测算法WSRFCM-LDOF.该算法采用集成粗糙集和阴影集的簇特征加权模糊聚类(WSRFCM)技术作为减少计算量的方法;簇特征加权的聚类算法可以有效处理分布不均匀的簇划分,在此基础上应用粗糙集和阴影集,使得簇特征加权模糊聚类算法可以有效划分交叠的簇,尤其对噪声和异常数据的处理具有高效性.实验表明,所提算法在降低时间复杂度的同时,提高了检测精度.
To reduce the amount of calculation for local outlier factor, on the basis of the LDOF algorithm, this paper proposed a novel outlier detect algorithm WSRFCM-LDOF. The algorithm adopted the integration of rough set and shadowed set into feature weighted fuzzy clustering, as a method of reducing the computational effort of local outliers. Associating feature with weights for each cluster is a common approach in clustering algorithms, and it can handle the different distribution of clusters effectively. The experimental results show the proposed algorithm has reduced the time complexity, meanwhile has improved the accuracy of detecting outliers.
出处
《微电子学与计算机》
CSCD
北大核心
2015年第10期129-133,共5页
Microelectronics & Computer
基金
国家自然科学基金项目(61203273
61202137)
南京信息工程大学本科生优秀毕业生论文(设计)支持计划
关键词
特征加权
阴影集
阴影粗糙模糊聚类
局部离群度
离群点检测
feature weights
shadowed set
shadowed rough-fuzzy clustering
local outlier degree
outlier detection
