基于BP神经网络的客户信用风险评价  

Customer credit risk assessment based on BP neural network

作　　者：于彤[1] 李海东[2] 

机构地区：[1]河海大学商学院,江苏南京211100 [2]北京建筑大学经济与管理工程学院,北京100044

出　　处：《现代电子技术》2014年第10期8-11,共4页Modern Electronics Technique

摘　　要：我国商业银行信用风险管理不足,已经严重影响银行的发展,因而银行需要重视客户信用风险评估。分析了银行信用风险的成因及评估存在的问题,从企业的财务情况出发,建立了客户信用风险评估指标体系。随机选取了我国制造业的160个上市公司样本,包括36个ST企业和124个非ST企业,并基于三层BP神经网络对样本进行训练及仿真测试,研究发现BP神经网络适用于信用风险评估,且其准确性优于Logistic回归模型。最后,从银行、企业、政府三个角度出发,对我国商业银行信用风险管理提出了一些建议及对策。The banks in China shouls pay more attention to the customer credit risk assessment because the commercial bank credit risk management in China is insufficient,which has seriously affected the development of banks. The formation cause of the bank credit risk and the problems existing in the assessment are analyzed. The customer credit risk assessment in-dex system was established on the basis of financial situation of enterprises. The 160 samples in listed companies in Chinese manufacturing industry were selected randomly,including 36 ST companies and 144 non ST companies,and then tested based on three-layer BP neural network training. It is found in the research that the BP neural network is suitable for the credit risk as-sessment,and its accuracy is better than that of Logistic regression model. Some suggestions and countermeasures to the credit risk management of Chinese commercial banks are put forward.

关 键 词：信用风险评估 评估指标体系 神经网络 商业银行 

分 类 号：TN911-34[电子电信—通信与信息系统] F832.1[电子电信—信息与通信工程]