一种鲁棒的局部异常检测算法  

A novel robust local outlier detection algorithm

作　　者：康欢[1] 张爱梅[1] 雷景生[2] 宋硕[2] 

机构地区：[1]贵州电网有限责任公司贵阳供电局,贵阳550004 [2]上海电力学院计算机科学与技术学院,上海200082

出　　处：《桂林电子科技大学学报》2017年第3期208-211,共4页Journal of Guilin University of Electronic Technology

基　　金：国家自然科学基金(61472236;61672337)

摘　　要：为解决异常检测算法的参数选择非常敏感问题,提出了一种鲁棒的局部异常检测算法。该算法用于任意形状和具有不同密度区域分布的数据集,能检测出局部离群点。实验结果表明,该算法的参数设置容易,对参数的选择不敏感,鲁棒性好。新算法用于电力用户电量异常检测,可帮助供电企业确定用户是否存在窃电行为。The anomaly detection algorithm is very sensitive to the parameter selection, a robust local outlier detection algo- rithm is proposed, This novel method not only can effectively identify global and local outliers but also is associated with one statistical parameter, Furthermore, the sole parameter can be easily determined without relying on users＇ domain knowl- edge, The method is insensitive to parameter variation, which ensures its superiority in robustness, In addition, the new algorithm is used to detect outliers of user electrical behavior, which can help power enterprises determine whether the user exists electricity stealing behavior.

关 键 词：异常检测 数据挖掘 鲁棒 用电行为 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]