一种基于高维空间聚类的离群数据发现算法  

Algorithm of Detecting Outliers Based on High-Dimension Space Clustering

作　　者：李旭辉[1] 郑丽英[1] 徐顼[1] 贾海鹏[1] 何知军[1] 

机构地区：[1]兰州交通大学电子与信息学院,甘肃兰州730070

出　　处：《微电子学与计算机》2007年第12期44-46,共3页Microelectronics & Computer

基　　金：教育部春晖计划(Z2004-1-6208);甘肃省自然科学基金项目(3ZS042-B2S-038)

摘　　要：针对高维稀疏聚类CABOSFV算法仅能解决二态变量高维稀疏聚类问题而对其他类型变量算法将失效的情况,通过定义"属性绝对值距离"解决了CABOSFV算法固有的这一缺陷。另外,针对聚类后产生伪孤立点问题,提出相应的异常数据处理策略,从而得到较为精确的离群数据。According to the high-dimension sparse clustering algorithm CABOSFV is only effective to solve the highdimension sparse clustering problems of two states variable. The ＂absolutely attribute distance＂ is defined, after thai, it solves the problems which CABOSFV can＇t solve, and according to the false outliers often happen after clustering, the paper gives some strategies to get better outlier data.

关 键 词：离群数据 CABOSFV算法 聚类 伪孤立点 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]