统计与聚类视角的中国中小企业信用评级研究

A Research on the credit rating to SMEs in China based on the perspective of statistics and clustering

基于回归分析的思想,可以对现有的聚类算法进行改进,以在一定程度上解决我国中小企业信用评级时面临的违约率缺失与主观性较高的问题。首先对影响回归、聚类以及二者间一致性的因素进行了分析,选取了与之相关的描述变量;接着根据这些变量构造蒙特卡洛模拟,将获得的模拟数据用于回归分析,得到了关于聚类结果与被解释变量间一致性和聚类最优输入参数两个回归方程;最后将回归方程用于指导聚类算法,对我国中小企业的信用风险进行了评级。结果显示,改进后的聚类结果与代表中小企业信用风险大小的企业存续期之间有较好的一致程度,同时还发现,资产类指标对中小企业信用风险的影响较为明显。

This paper points out that the existing clustering algorithm can be improved based on the idea of regression analysis to help overcome the problem of PD missing and high subjectivity concerning the credit rating to SMEs. First, we analyze the factors that can affect the result of regression, clustering and the consistency between them, and we choose the variables corresponding to these factors; Secondly, we build the Monte Carlo model using these variables, and through the regression analysis of the analog data, we obtain two regression equations about the consistency and the best input parameter. Finally, we use the two regression equations to guide the clustering algorithm and rate the credit risks of SMEs of China. The result shows that, the improved clustering result has a high consistency with the continuance which represents the credit risk of SMEs, and we also find that the asset class index has an obvious influence on the credit risk of SMEs.