摘要
为实现对电气故障快速、准确和动态的分类,提出一种有机结合实例和属性加权的朴素贝叶斯电气故障分类方法(AIWNB);朴素贝叶斯分类方法中的先验概率和条件概率采用两种实例加权方式加以改进,积极实例权值取决于各属性值频度的统计值,而消极实例权值通过逐条计算训练实例与测试实例间的相关性加以确定;属性权值则基于互信息定义为属性-属性相关性和属性-类相关性之间的残差;所提出的AIWNB方法将属性加权和实例加权有机结合在朴素贝叶斯统一框架内,利用高低压用户的电气实测数据进行验证,实验结果表明,与朴素贝叶斯相比,加权后的朴素贝叶斯方法更具竞争性,准确率和F1分数可提升3.09%和9.39%,证明所提的AIWNB算法在电气故障分类的实用性及有效性;并与其他电气故障分类方法进行对比,验证算法的优越性。
In order to achieve rapid,accurate and dynamic classification of electrical faults,an electrical fault classification model based on attribute and instance weighted naive bayes(AIWNB)is proposed.The prior probability and conditional probability in the naive Bayes classification method are improved by using two instance weighting methods.The eager instance weight depends on the statistics of the frequency of each attribute value,and the lazy instance weight is determined by calculating the correlation between the training instance and the test instance one by one.Attribute weight is defined as the residual between attribute-attribute correlation and attribute-class correlation based on mutual information.The proposed AIWNB method organically combines attribute weighting and instance weighting in a unified framework of Naive Bayes,the electrical measurement data of high and low voltage users is used to verify.Experimental results show that compared with pure Naive Bayes,the weighted Naive Bayes is more competitive,and the accuracy and F1 score can be increased by 3.09% and 9.39%,the practicality and effectiveness of algorithm are proved in the classification of electrical faults.Compared with other electrical fault classification methods,the superiority of algorithm is verified.
作者
舒一飞
郭汶昇
樊博
康洁滢
许诗雨
杨林
SHU Yifei;GUO Wensheng;FAN Bo;KANG Jieying;XU Shiyu;YANG Lin(Marketing Service Center,State Grid Ningxia Electric Power Co.,Ltd.,Yinchuan 750000,China;College of Electrical Engineering,Sichuan University,Chengdu 610065,China)
出处
《计算机测量与控制》
2022年第5期169-174,180,共7页
Computer Measurement &Control
基金
国网宁夏电力有限公司营销服务中心(国网宁夏电力有限公司计量中心)科研项目(JG29YX210027)。
关键词
属性加权
积极实例加权
消极实例加权
朴素贝叶斯框架
电气故障分类
attribute weighted
eager instance weighted
lazy instance weight
naive bayes framework
electrical faults classification
