基于主成分分析法和BP神经网络的电力客户信用评价  

Evaluation on the Electricity Customers' Credit Based on PCA and BP Neural Network

作　　者：宋新明[1,2] 居勇[2] 曾鸣[2] 卫炜[2] 褚烨[2] 

机构地区：[1]广东电网公司阳江供电局,广东阳江529500 [2]华北电力大学工商管理学院,北京102206

出　　处：《技术经济与管理研究》2009年第5期22-24,共3页Journal of Technical Economics & Management

摘　　要：为了加强电力公司对电力客户信用风险的事先控制,降低电力公司运营风险,需要对电力客户按信用等级进行分类,确定不同客户的信用风险等级,以便执行不同的信用风险控制策略。本文通过将定量与定性的指标相结合,建立了电力客户信用评价指标体系。引入主成分分析法和改进的BP神经网络法,将两者相结合,建立数学评价模型。通过将指标体系中的各指标带入该模型进行测算,可以计算出电力客户信用风险大小,从而确定各个客户信用风险等级。实例研究表明,利用此指标体系和数学模型能够准确地判断电力客户所处的信用等级,对于电力公司规避电力客户信用风险有较强的实际指导意义。A research is given to determine the credit grade of electricity customers and classify the customers into different risk levels in order to enhance the prior controlling work for the power supplying company and to reduce the operating risk. First, a risk evaluation index system is set up including both quantitative and qualitative indexes. Second, principal component analysis method （PCA） and improved BP neural network method are introduced. A model to evaluate the credit risk of electricity customers is established by combining these two methods together. Then each index of the index system is put into the model, so the risk level of every customer can be determined. At last, a test is made to demonstrate the effectiveness of the index system and the mathematic model. The result shows that this method can be adopted in the credit risk assessment since it could give an accurate judgment of risk management.

关 键 词：电力客户 信用风险 主成分分析 BP神经网络 

分 类 号：F407.61[经济管理—产业经济]