异常点挖掘研究进展  

Research progress on outlier mining

作　　者：王宏鼎[1] 童云海[1] 谭少华[1] 唐世渭[1] 杨冬青[1] 

机构地区：[1]北京大学视觉与听觉处理国家重点实验室,北京100871

出　　处：《智能系统学报》2006年第1期67-73,共7页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金资助项目(60403041).

摘　　要：异常点是数据集中与其他数据显著不同的数据.一个人的噪声对另一个人而言可能是有用的数据,因此,随着人们对数据质量、欺诈检测、网络入侵、故障诊断、自动军事侦察等问题的关注,异常点挖掘在信息科学研究领域日益受到重视.在充分调研国内外异常点挖掘研究文献基础上,系统地综述了数据库研究领域中异常点挖掘的研究现状,对已有各种异常点挖掘方法进行了总结和比较,并结合当前研究热点,展望了异常点挖掘未来的研究方向及其面临的挑战.An outlier is a data point that is significantly different from the others in a data set. One person＇s noise could be another person＇s signal, and therefore the problem of outlier mining attracts more and more interests in research of information science when the research fields of data quality, fraud detection, intrusion detection, fault diagnosis, military scout and so on receive wide attentions. In this paper, a survey was presented for the problem of outlier mining from the basic concepts to the principal research problems and the underlying techniques, including origination of outlier, definition of outlier and the comparison of popular outlier mining methods. A summary of the current state of the art of these techniques, a discussion on future research topics, and the challenges of the outlier mining were also presented.

关 键 词：异常点 挖掘方法 局部异常点 数据流 高维数据 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]