基于孤立森林算法的供电营销大数据异常挖掘方法  

Anomaly Mining Method of Power Supply Marketing Big Data Based on Isolated Forest Algorithm

作　　者：于亮[1] 钟宏伟[1] 冯祎辰 肖莞 张硕 YU Liang;ZHONG Hongwei;FENG Yichen;XIAO Wan;ZHANG Shuo(Customer Service Center of State Grid Beijing Electric Power Company,Beijing 100078,China)

机构地区：[1]国网北京市电力公司客户服务中心,北京100078

出　　处：《微型电脑应用》2025年第2期51-54,共4页Microcomputer Applications

基　　金：国网北京市电力公司科技项目(kymg202212)。

摘　　要：当前供电营销大数据异常挖掘方法中异常数据检测环节较为薄弱,导致数据异常挖掘率相对较低且误检率较高,为此,提出基于孤立森林算法的供电营销大数据异常挖掘方法。构建孤立森林模型检测原始数据,借鉴二叉搜索树(BST)方法确定模型之间数据距离,获取异常数据。使用离散小波变换方法,提取异常数据特征。对传统k-means算法进行优化,结合异常数据特征构建数据簇,实现对异常数据的挖掘分析。构建应用测试环节,测试结果表明,所提方法可有效提升供电营销大数据异常挖掘率,进一步降低数据异常挖掘误检率,为供电营销大数据分析环节提供新的发展方向。The detection link of abnormal data in the current power supply marketing big data anomaly mining method is relatively weak,which leads to the relatively low data anomaly mining rate and high false detection rate.Therefore,a power supply marketing big data anomaly mining method based on isolated forest algorithm is proposed.The isolated forest model is constructed to detect the original data,and the binary search tree(BST)method is used for reference to determine the data distance between models to obtain abnormal data.Discrete wavelet transform method is used to extract the features of abnormal data.The traditional k-means algorithm is optimized,and data clusters are constructed based on the features of abnormal data to realize the mining and analysis of abnormal data.An application test link is built,and the test results show that this method can effectively improve the anomaly mining rate of power supply marketing big data,further reduce the false detection rate of data anomaly mining,and provide a new development direction for the power supply marketing big data analysis link.

关 键 词：孤立森林算法 供电营销 大数据平台 异常数据挖掘 聚类数目 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]