基于二阶段Logistic回归的小企业信用指标判别模型  

Credit Index Discrimination Model of Small Enterprises Based on Two-Stage Logistic Regression

作　　者：郭林 李战江[1] 孔文婷 GUO Lin;LI Zhan-jiang;KONG Wen-ting(School of Economics and Management,Inner Mongolia Agricultural University,Huhhot 010010,China)

机构地区：[1]内蒙古农业大学经济管理学院,内蒙古呼和浩特010010

出　　处：《数学的实践与认识》2020年第20期35-45,共11页Mathematics in Practice and Theory

基　　金：国家自然科学基金重点项目(71731003);内蒙古自治区自然科学基金(2020MS07009);内蒙古自治区科技计划项目(201605053);内蒙古农业大学博士后科学基金后期资助项目。

摘　　要：现有信用指标遴选大多仅仅对客户是否违约进行判别,而忽略了高违约损失与低违约损失是一类性质不同的信用特征问题.通过将是否违约与不同违约损失相结合,提出了两阶段判别的信用评价指标组合筛选新思路,构建了敏感性更强、判别力更大的信用评价指标体系,并使用3111个小企业的信贷数据进行了实证分析.基于二阶段Logistic回归的小企业信用指标判别模型是指首先通过Logistic回归将客户分为违约与不违约进行第一阶段判别,然后利用样品有序聚类将违约客户分为高违约损失率客户与低违约损失率客户,最后通过Logistic回归对高违约损失率客户和低违约损失率客户进行第二阶段判别.结果显示一是成本费用净利率等九个指标是可以区别不违约、低违约、高违约这三类不同违约特征的具有高判别力的小企业信用评价指标.二是ROC曲线的有效性检验表明基于二阶段Logistic回归的小企业信用指标判别模型得到的指标体系是有效的.Most of the existing credit index selection only discriminates whether the customer is in default but ignores that high default loss and low default loss are a kind of credit characteristic problem with different nature.By combining whether or not default with different default losses,this paper puts forward a new idea of two-stage discriminant combination screening of credit evaluation indicators,constructs a more sensitive and discriminant credit evaluation index system,and uses the credit data of 3,111 small enterprises for empirical analysis.The small enterpeise credit index discrimination model based on two-stage Logistic regression means that customers are divided into default and non-default through Logistic regression for the first stage discrimination.Then,the sample orderly clustering is used to divide the default customers into high default loss rate customers and low default loss rate customers.Finally,Logistic regression is used to distinguish customers with high default loss rate and customers with low default loss rate in the second stage.The results show that the nine indicators,such as the net interest rate on costs,are the credit evaluation indicators of small enterprises with high discriminant power that can distinguish three types of default characteristics:non-default,low default and high default.Second,the effectiveness test of ROC curve shows that the indicator system obtained based on the two-stage Logistic regression model of small enterprise credit indicator discrimination is effective.

关 键 词：信用评价 指标筛选 二阶段Logistic回归 小企业 

分 类 号：F276.3[经济管理—企业管理] F832.4[经济管理—国民经济] O212.1[理学—概率论与数理统计]