商业银行信用风险评估实证分析及方法比较  

Empirical Analysis and Comparison on Methods of Credit Risk Evaluation for Commercial Banks

作　　者：李君艺[1] 张宇华[2] 

机构地区：[1]东莞职业技术学院计算机工程系,广东523808 [2]广东工业大学计算机学院,广东510006

出　　处：《网络安全技术与应用》2013年第3期76-79,共4页Network Security Technology & Application

摘　　要：数据挖掘技术为商业银行信用风险管理问题提供了新的思路和方法。本文运用三种常用的数据挖掘方法——多元判别分析、聚类分析及贝叶斯网络模型,以商业银行的客户信用风险评级指标数据为样本,对信用风险评估方法进行实证分析,对三种方法的验证结果进行比较。结论表明,在信用风险各项属性指标之间条件相互依赖的情况下,贝叶斯网络模型优于其它两种方法。Data mining technology provides a new idea and method for the credit risk management of commercial banks. This paper compares three kinds of common data mining methods -- multiple discriminant analysis, cluster analysis and the Bayesian network model, which are applied in credit risk evaluation of commercial banks, using the real data sample of a national commercial bank.The comparison result shows that, under the condition of mutual dependence among the credit risk indicators, Bayesian network model is better than the other two methods.

关 键 词：信用风险 多元判别分析 聚类分析 贝叶斯网络 

分 类 号：F832.33[经济管理—金融学] F224