关键词: 数据挖掘, 常规算法, 离群检测, 应用

Abstract: 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 outliers.Outlier 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.

Key words: data mining, regular algorithm, outlier detection, application