软计算信用分析方法比较探讨  

A Comparison Research of Soft Computing Applied to Credit Scoring

作　　者：梁 吴德胜[1] 

机构地区：[1]中国科学技术大学商学院,合肥230026

出　　处：《系统工程理论方法应用》2005年第3期268-274,共7页Systems Engineering Theory·Methodology·Applications

基　　金：国家863资助项目(2002AA41361)

摘　　要：研究了信用分析中的软计算方法——BP网络与自适应神经模糊推理,并对软计算方法与DEA、TOPSIS的结合运用作了进一步探讨。采用V-foldCross-validation技巧,利用样本公司实际指标数据对该模型的评分效果进行了实证研究。Soft computing including BP neural networks and Adaptive Neural Fuzzy Inference applied to Credit scoring is studied in the current paper. Furthermore, the combination of Soft computing and DEA as well as TOPSIS is discussed. Finally, experiential research of the credit-score models' scoring powers is carried out by use of experiential data and by V-fold Cross-validation technique.

关 键 词：信用评分 自适应神经模糊推理 BP神经网络 V-fold Cross-validation技巧 软计算 数据包络分析 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]