常规挖掘算法在离群数据检测中的应用  

Application of regular data mining algorithms in outlier detection

作　　者：金义富[1] 朱庆生[2] 

机构地区：[1]湛江师范学院信息学院,广东湛江524048 [2]重庆大学计算机学院,重庆400044

出　　处：《计算机工程与应用》2008年第16期145-147,共3页Computer Engineering and Applications

基　　金：重庆市自然科学基金(the Natural Science Foundation of Chongqing City of China under Grant No.2005BB2224)

摘　　要：数据挖掘以发现常规模式为主体,但离群数据在欺诈分析及安全领域具有重要分析价值,离群数据检测已成为数据挖掘的重要内容。对聚类与分类以及关联规则分析中典型的常规数据挖掘算法如何处理离群数据进行全面分析与总结,讨论了BIRCH、CURE、Chameleon、DBSCAN以及基于共享最近邻的聚类算法以及基于不平衡分类和基于非频繁模式的离群检测技术,给出了一种利用K-最近邻算法的离群数据检测方法,并报告了测试结果。In general,data mining is mainly discovering for regular patterns.It is an important part of data mining to detect outliers as the significance of analyzing for outliers is great in fraud analysis and security fields.This paper analyzes and summarizes roundly typical regular data mining algorithms in clustering,classification and association rules how to deal with outliersOutlier detection methods based on unbalanced classification,stering algorithms such as BIRCH,CURE,Chameleon, DBSCAN and shared near neighbour are discussed mainly.An outlier detection algorithm based K-near neighbour is put forward in the paper,and its test result is reported.

关 键 词：数据挖掘 常规算法 离群检测 应用 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]