基于kappa融合系数排序选择性集成的窃电检测  

Electricity theft detection based on kappa fusion coefficient sorting selective integration

作　　者：黄冬梅 王子豪 胡伟 胡安铎 孙锦中 孙园 HUANG Dongmei;WANG Zihao;HU Wei;HU Anduo;SUN Jinzhong;SUN Yuan(College of Electronic and Information Engineering,Shanghai University of Electric Power,Shanghai 201306,China;College of Economics and Management,Shanghai University of Electric Power,Shanghai 201306,China;College of Mathematics and Physics,Shanghai University of Electric Power,Shanghai 201306,China)

机构地区：[1]上海电力大学电子与信息工程学院,上海201306 [2]上海电力大学经济管理学院,上海201306 [3]上海电力大学数理学院,上海201306

出　　处：《电子设计工程》2025年第7期90-94,共5页Electronic Design Engineering

摘　　要：针对传统窃电检测中分类检测模型的单一性和局限性,提出基于kappa融合系数排序选择性集成窃电检测方法。通过对七个分类器进行训练,计算出各项指标,将分类器依据误检率以及kappa进行融合,计算融合系数进行排序,筛选出性能优越的基分类器。根据误差分解,计算基分类器误差与方差,衡量预测值与实际值之间的差距,赋予基分类器权重进行加权投票集成。实验结果表明,对比传统的窃电检测模型,所提模型在多项评价指标下表现较好,具有良好的检测性能。In view of the singularity and limitations of the classification detection model in traditional electricity theft detection,a selective integrated theft detection method based on kappa fusion coefficient sorting is proposed.Through the training of seven classifiers,the indicators is calculated.Each classifier is fused according to the false detection rate and kappa.Sorting the classifiers is based on fusion coefficient to screened out superior performance of classifiers.Based on the error decomposition,the error and variance of each classifier are calculated in order to measure the gap between the predicted value and the actual value.Finally,the individual classifier weights are assigned for weighted voting integration.Experimental results show that the proposed model performs better under multiple evaluation indexes and has good detection performance compared with the traditional theft detection model.

关 键 词：窃电检测 KAPPA 选择性集成 误差分解 

分 类 号：TN0[电子电信—物理电子学]