基于Fisher判别的小型工业企业债信评级模型及实证

Small Industrial Enterprises' Credit Rating Model and Empirical Analysis Based on Fisher Discriminant

本文以中国某商业银行1814个小型工业企业贷款客户数据为样本,根据指标对违约状态鉴别精度的影响程度进行第一次筛选,保证遴选出的指标对违约状态鉴别能力都有显著影响;根据准则内相关分析进行第二次筛选,避免遴选出的指标反映信息重复,构建了一套能显著区分小型工业企业违约状态的评级体系。本文的创新与特色:一是根据有、无特定指标两种状态下、Fisher判别对违约状态鉴别精度的提高或降低,反映特定指标对违约状态的影响程度,剔除对违约状态的判别精度没有影响或有降低影响的指标,保留可以显著提高违约状态判别精度的指标,完善了现有研究遴选指标的标准与违约状态无关的不足。二是在相关系数大于0.7的两个指标中,根据对Fisher判别精度影响程度越大、这个指标区分违约状态能力越强的思路剔除对Fisher判别精度影响程度较小的指标,避免了现有研究在剔除冗余指标时、对违约状态影响大的指标可能被误删的不足。

According to the influence degree of indexes on the identification precision of default state,this study conducts the first screening to ensure that the selected indexes have a significant effect to the identification ability of default state. According to correlation analysis within guidelines,this study conducts the second screening to keep the selected indexes from reflecting the same information.This study builds a set of rating system which can significantly distinguish the default state of small industrial enterprises. This paper makes an innovative and characteristic attempt in two aspects. Firstly,according to the condition either with or without specific indicators,effects of using Fisher discriminant on identification accuracy of default state,it reflects the influence of specific indicators on the default state. Secondly,with regards to the two indicators with correlation coefficient above 0.7,this paper,based on the thought that the greater impact on the Fisher discriminant accuracy,the stronger ability of the index to distinguish default state,accurately removes the indexes that have smaller influence on Fisher discriminant.