统计研究 ›› 2008, Vol. 25 ›› Issue (7): 61-65.

• 论文 • 上一篇    下一篇

基于神经网络的贷款违约识别研究

刘 弘   

  1. 上海财经大学
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2008-07-15 发布日期:2008-07-15

A Research on Default Loan Recognition Based on Neural Networks

Liu Hong   

  • Received:1900-01-01 Revised:1900-01-01 Online:2008-07-15 Published:2008-07-15

摘要: 随着国内金融市场的逐步开放,我国商业银行将面临日益激烈的市场竞争。提高商业银行的信用风险管理水平,增强市场竞争力,已经成为非常迫切的问题。除了完善信贷管理体制以外,研发信用风险模型以降低信用风险很有必要。鉴于目前的数据情况,现实的选择是研发贷款违约识别系统。本文对神经网络用于贷款违约识别做了实证研究,并与判别分析和决策树做了性能比较,得到了一些有意义的结论。

关键词: 违约识别, 信用风险, 神经网络, 特征选择

Abstract: As domestic financial market is open gradually, Chinese commercial banks are going to facing increasing competition. It’s urgent to improve management ability of credit risk to increase their competitiveness. Besides perfecting the management system of loan, it’s necessary to develop credit risk model for the decrease of default risk. According to the quality of domestic credit data, the most practical choice is to develop default loan recognition models. Compared with discriminant analysis and decision tree, the performance of neural networks is mainly studied and some meaningful conclusions are drawn.

 

Key words: Default recognition, Credit risk, Neural networks, Feature selection