摘要
针对某电力系统的用户窃漏电行为进行数据挖掘,利用LM神经网络算法建立评价模型。首先对预处理后的数据进行数据变换,得到用来表征用户窃漏电行为的评价指标,其次基于LM算法建立了用户窃漏电行为预测的评价模型,实现对用户窃漏电行为的监测,并且与分类回归树算法(CART)进行对比,利用ROC曲线评价方法进行评估,用随机森林(RF)和支持向量机(SVM)算法进行了实验验证,实验结果表明,相对于CART、RF、SVM算法,基于LM神经网络的用户窃漏电检测系统得到更高得预测准确率。
A Levenberg-Marquardt(LM)neural network method is used to establish a model for data mining in which users are aware of electrical leakage behaviors.Lagrange interpolation method was used to interpolate the original data,and data was transformed to obtain new evaluation indicators to characterize the law of burglary leakage behavior.Based on the LM neural network,a comprehensive evaluation model of user burglary leakage behavior was established to achieve burglary.The effective recognition of the behavior is compared with the CART decision tree algorithm.The ROC curve evaluation method is used for evaluation,and the experimental results are verified by RF and SVM algorithms.The experimental results show that compared with the CART、RF、SVM algorithm,the user sniffer leakage detection system based on LM neural network obtains higher prediction accuracy.
作者
赵文仓
陈聪聪
郑鸿磊
Zhao Wencang;Chen Congcong;Zheng Honglei(College of Automation and Electronics Engineering,Qingdao University of Science & Technology,Qingdao 266061,China)
出处
《电子测量技术》
2018年第24期119-122,共4页
Electronic Measurement Technology
基金
国家自然科学基金(61171131)
山东省重点研发计划(2013YD01033)项目资助
