基于偏差熵的可持续用电客户信用综合评级模型

A Credit Rating Model of Sustainable Power Customers Based on Deviation Entropy

可持续用电客户的信用评级是电力企业规避欠缴电费风险的有效手段,文章以提升电力企业预防拖欠电费风险的能力,实现各信用级别客户的差别化管理和服务为目的,构建基于偏差熵的可持续用电客户信用综合评级模型。利用综合评价方法评价用电客户信用等级,借助偏差熵量化描述综合评级方法中所有单一评级方法的相容性质量,确定综合评级和单一评级结果之间的误差,保证综合偏差最小化;利用拉格朗日乘子法求解综合评级方法内相容性质量最高的权重项,量化综合评级方法自身的相容性,既考量到各单一评级方法和综合评级方法之间的综合偏差,也考量到所有评级方法之间的相容性质,通过客户信用综合评价值计算模型最终确定用电客户的信用等级。实验结果表明,该模型正确率高达99.7%,与专家组标准结果基本一致,准确率为94.8%,召回率为85.6%,能够客观、合理、有效地实现信用评级,为用电客户信用评估提供参考。

The credit rating of sustainable electricity customers is an effective means for power enterprises to avoid the risk of arrears of electricity charges.In order to improve the ability of power enterprises to prevent the risk of arrears of electricity charges and realize the differentiated management and service of customers with various credit levels,a comprehensive credit rating model of sustainable electricity customers based on deviation entropy is constructed.The comprehensive evaluation method is used to evaluate the credit rating of electricity customers,and the compatibility quality of all single rating methods in the comprehensive rating method is quantitatively described with the help of deviation entropy,so as to determine the error between the comprehensive rating and the single rating results,and minimize the comprehensive deviation.Lagrange multiplier method is used to solve the weight terms with the highest compatibility quality in the comprehensive rating method,and the compatibility of the comprehensive rating method itself is quantified.The comprehensive deviation between each single rating method and the comprehensive rating method is considered,and the compatibility between all rating methods is also considered.Through the calculation model of customer credit comprehensive evaluation value,the credit rating of electricity customers is finally determined.Experimental results show that the correct rate of this model is up to 99.7%,which is basically consistent with the standard results of the expert group,with an accuracy of 94.8%and a recall rate of 85.6%,which can realize the credit rating objectively,reasonably and effectively and provide reference for the credit evaluation of electricity customers.