基于V-foldCross-validation和Elman神经网络的信用评价研究  

Research of Credit Score Based on V-fold Cross-validation and Elman Neural Networks

作　　者：吴德胜[1] 梁樑[1] 

机构地区：[1]中国科学技术大学商学院,安徽合肥230052

出　　处：《系统工程理论与实践》2004年第4期92-98,共7页Systems Engineering-Theory & Practice

基　　金：国家 863项目基金(2002AA41361)

摘　　要：　研究了关于公司信用评估问题的现状,指出一般神经网络应用于信用评估领域的不足.在此基础上,提出一套甄选原则以选择关键的信用评分指标;然后依据这些指标建立了基于Elman回归神经网络的我国企业的信用评估模型.采用V-foldCross-validation技巧对该模型的评分效果进行了实证研究.In this paper we first point out the gaps of associated research by reviewing the corporate credit scoring research by general neural networks. Then some practical principles are carefully considered so that the distinguished measures can be selected. These measures are used to develop credit-scoring models applied to Chinese corporations based on Elman neural networks. Finally, V-fold Cross-validation technique is applied to test the model's scoring power.

关 键 词：ELMAN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]