基于朴素贝叶斯分类器的电力系统信誉评级研究

Credit Evaluation of Power Consumers Based on Naive Bayesian Classifier

目前电力企业普遍面临的一个难题是客户的巨额欠费。为此,电力公司每年都需要投入大量的精力进行电费催缴工作,但问题至今仍难以得到根本解决。通过从电力系统大规模的缴费数据中提取训练样本,利用朴素贝叶斯算法进行学习,得到分类规则,将这些规则应用于对用电客户的信誉进行评级,建立客户信誉评级系统,可以帮助电力企业更好地管理客户缴费行为,并为其运营管理提供数据支持。结果表明,对于不同的分类样本甚至不同的指标体系,都可以获得较好的分类效果。

To collect huge amount of outstanding balance owned by power consumers is currently one of the toughest jobs for power suppliers.Up till now,no effective solutions have been found though enormous efforts are made to tackle it every year.To deal with this dilemma,this paper is to design a credit evaluation system to evaluate the credit levels of power consumers.For this end,samples are selected from the massive data of power bill payment and the Naive Bayesian Algorithm is employed to find out the working principles of classification.All the empirical researches on different groups of samples and index systems reveal that the evaluation system is effective and efficient in managing power bill payment and providing statistical support for the operation of power suppliers.