基于特征赋权的离群数据再聚类算法  

A Re-clustering Algorithm Based on Feature Weighting of Outlier Data

作　　者：李旭辉[1] 郑丽英[1] 

机构地区：[1]兰州交通大学电子与信息工程学院,甘肃兰州730070

出　　处：《兰州交通大学学报》2008年第1期135-137,共3页Journal of Lanzhou Jiaotong University

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

摘　　要：离群数据发现与分析是数据挖掘的重要组成部分,现有离群数据挖掘算法主要针对如何检测离群对象,缺乏对挖掘出的离群数据集进行解释与分析的有效方法.通过对离群数据来源及特性进行分析,定义了离群贡献度的概念,提出了一种基于特征赋权的离群数据再聚类算法.实验表明,该算法可以有效地揭示离群数据产生来源,有助于对整体数据集的更全面了解.It is an important part of data mining to discover and analyze outlying observations. Existing out- lier mining algorithm focuses on detecting and identifying outliers, but studies of outlier include both mining outliers and analyzing why they are exceptional. By analyzing the origin and feature of outliers, a concept of exceptional contribution degree is defined and then an algorithm for re-clustering outliers based on feature weighting is proposed. Experimental results show that the approach can be used for identifying the origin of outliers and improve the understanding of whole data set.

关 键 词：离群点 关键域子空间 离群贡献度 K-MEANS算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]