改进的局部异常因子算法用户窃电检测研究
Research on improved local outlier factor algorithm for user electricity theft detection
投稿时间:2022-04-02  修订日期:2022-04-02
DOI:
中文关键词: 窃电检测  改进LOF  球树模型
英文关键词: electricity theft detection  improved LOF  ball tree model
基金项目:国家社会科学基金项目-重大招标项目(19ZDA284);四川省科技项目(2020JDR0141, 2020JDRC0040);中央高校优秀学生培养工程项目(2021NYYXS119)
作者单位邮编
殷锋 西南民族大学 
周绍军 四川水利职业技术学院 430074
漆翔宇 西南民族大学 
曹旭 西南民族大学 
摘要点击次数: 91
全文下载次数: 0
中文摘要:
      异常值检测作为数据挖掘领域研究的热点问题之一,广泛应用于窃电识别、反信息欺诈等领域。而LOF算法作为一种依赖数据密度进行异常值识别的算法,因其具有检测精度高、应用场景多元等优势常被应用于窃电识别与检测过程中,但该算法往往存在较高的时间复杂度。针对该问题,提出基于混合剪枝树模型改进的RBT-LOF算法,并在此基础上提出了相应的窃电用户识别模型。RBT-LOF算法首先对混合剪枝树的超平面划分方式进行调整,采用数据第一特征向量找出平衡分割位并重构数据对象;其次使用混合剪枝查询加速数据对象的搜索。实验结果表明,基于RBT-LOF的窃电识别模型较LOF算法、SVM、CNN和WDNet模型具有更高的执行效率和检测精确率。
英文摘要:
      As one of the hot issues in the field of data mining, outlier detection is widely used in the fields of electricity theft identification, anti information fraud and so on. LOF algorithm, as an algorithm for outlier identification based on data density, is often used in the process of power theft identification and detection because of its advantages of high detection accuracy and multiple application scenarios, but the algorithm often has high time complexity. To solve this problem, an improved RBT-LOF algorithm based on hybrid pruning tree model is proposed, and the corresponding power stealing user identification model is proposed. Firstly, the RBT-LOF algorithm adjusts the hyperplane division mode of the hybrid pruning tree, uses the first feature vector of the data to find the balanced segmentation bit and reconstruct the data object; Secondly, hybrid pruning query is used to speed up the search of data objects. The experimental results show that the electricity theft identification model based on RBT-LOF has higher execution efficiency and detection accuracy than LOF algorithm, SVM, CNN and WDNet model.
View Fulltext   查看/发表评论  下载PDF阅读器
关闭