不同模型在信用评价中的比较研究  

Comparison Research of Different Methods Applying to Corporate Credit Score

作　　者：吴德胜[1] 梁樑[1] 杨力[1] 

机构地区：[1]中国科学技术大学商学院,安徽合肥230026

出　　处：《预测》2004年第2期73-76,69,共5页Forecasting

基　　金：国家863基金资助项目(2002AA41361)

摘　　要：比较了不同模型应用于企业信用评价问题的优劣,针对信用评分问题特点,采用Elman回归神经网络和BP网络建模。在建立了适合于我国企业的信用评分指标体系之后,运用以上两种方法进行实证研究并比较两种网络的诊断行为;为克服小样本建模的缺点,引进V foldCross validation计算技巧。Different methods applying to corporate credit score are studied. By use of Elman neural networks and BP neural networks, credit-score models are developed combining the characteristic of Chinese corporate credit score. After some distinguished indexes are chosen, the two methods are comparatively studied by use of experiential data. The V-fold Cross-validation technique is supposed by the authors to overcome the shortage of the small sample.

关 键 词：ELMAN神经网络 BP神经网络 V-fold Cross-validation技巧 信用评分 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]