一个基于小样本的银行信用风险评级模型的设计及应用  

A Credit Rating Model for Analyzing Bank Customers Based on Small Sample

作　　者：迟国泰[1] 潘明道[1] 齐菲[1] 

机构地区：[1]大连理工大学工商管理学院

出　　处：《数量经济技术经济研究》2014年第6期102-116,共15页Journal of Quantitative & Technological Economics

基　　金：国家自然科学基金项目(71171031);国家自然科学基金青年科学基金项目(71201018);教育部科学技术研究项目(2011-10);教育部人文社会科学研究青年基金项目(11YJC790157);中国银监会银行业信息科技风险管理项目(2012-4-005);河北省自然科学基金青年科学基金(G2012501013);中国邮政储蓄银行总行小额贷款信用风险评价与贷款定价资助项目(2009-07);大连银行小企业信用风险评级系统与贷款定价项目(2012-01)的资助

摘　　要：银行信用风险评价用于判别不同银行风险的大小。通过组合赋权,可以确定银行信用风险评级指标的权重。运用评价得分的分布检验找到得分的分布规律,进而模拟、扩充样本数据,根据扩充数据划分银行的信用等级。通过对评价得分分布检验,可以揭示银行信用评价得分服从对数分布的规律,解决样本数据有限的条件下如何进行数据扩充的难题;根据对数分布规律进行数值模拟,对符合分布规律扩充1000倍后的得分数据进行信用等级划分,能够解决小样本无法进行信用评级的难题。实证结果表明,评级结果的序关系与权威机构评级结果的序关系一致。Credit risk rating is used to judge the size of different bank risks. The weight of credit rating index is determined by combination weighing approach. This paper tests the distribution of rating scores and finds out their distribution rules, from which data are generated and sample date are expanded. In the end, the ranks of bank credit risks can be divided. The innovations and features of this paper are as follows： through testing the distribution rule of rating scores, the logarithm distribution rule of rating scores on credit risk evaluation of commercial bank is re- vealed. This solves the difficult problem how to expand the small sample data while the sample is very limited in the real world. Through the numerical simulation on the logarithm distribution rule, the credit ratings score grading are gotten after 1000 times the data expansion of the original sample under the distribution rule. This solves the problem which the small sample can＇t be credit grading.

关 键 词：信用风险 银行评级 样本检验 样本扩充 

分 类 号：F830.33[经济管理—金融学]