摘 要
中国的商业银行开展客户评级、贷款评级的的时间不是太长，数量也不多，
目前还没有一套科学的成熟可靠的信用评级体系。各以工农中建四大国有商业
银行为首的中国商业银行正在摸索和建设自己的内部评级体系。但是我国目前
既没有建立一个比较权威性的第三方评估机构，又缺乏一套科学而有效的评级
体系。没有制定行业内通行的统一规范，各家商业银行推行自己的客户评级标
准。同一公司客户在不同银行的信用评级结果差异颇大。而国外发达国家，既
有正规的第三方评级机构，也有可靠而有效的评级体系。国内商业银行的信用
风险管理水平和意识还有待加强和提高。由于信用评级结果的可靠性取决于参
与评级客户所提供数据的真实性，故根据银行内部信贷数据来构建一套客户信
用风险评级体系显得尤为重要。
本文首先介绍了信用风险的定义、新巴塞尔协议下的内部评级法（IRB）
的主要内容。对比分析了国内外对违约概率模型 PD 的研究情况。接着介绍了
工商银行湖南分行内部信用评级工作的开展情况，对比分析了工商银行内部评
级一期和二期项目所取得的成就及不足之处。最后运用 Logistic 回归模型对湖
南工行所属的企业客户进行实证分析，构建湖南工行企业信用风险评估模型。
在模型的构建时，随机选取了 72 家企业作为建模样本组，随机选取了 58 家企
业作为检验样本组，通过处理样本企业 2012 年的数据得到相应的 40 个财务指
标，利用 SPSS 软件进行正态性检验、假设检验、相关性检验来筛选指标，最
后再利用 Logistic 回归模型对筛选出来的指标进行回归分析。
通过使用检验样本组对该模型进行验证，发现该模型的准确率达到了
84.5%。本文所得出的公司客户信用评价模型对于湖南工行现有的情况来说是
较为有效的。
关键词：工商银行湖南分行；内部信用评级；Logistic 回归模型；信用风险
Abstract
The time of customer rating and loan grading carried out by China's
commercial banks is not too long, and the quantity is not much. Currently, there is
not a scientific and reliable credit rating system. Led by four state -owned
commercial banks, the commercial banks in china are groping and establishing their
own internal rating system. But in China, there has neither authoritative third-party
assessment agency, nor scientific and effective rating system. Without unified
industry standards, the commercial banks implement their own customer rating
criteria. The differences of results of credit rating in different banks are significant.
In developed countries, however, not only have formal third-party rating agencies,
but also have reliable and effective rating system. The level and awareness of credit
risk management in domestic commercial banks have yet to be strengthened and
enhanced. Because the reliability of credit rating depending on the authenticity of
data provided by rating customers, therefore, building a set of customer credit risk
rating system according to the bank's internal credit data is particularly important.
In this paper, we firstly introduced the definition of credit risk and the main
content of internal ratings-based (IRB) under the New Basel Accord, analyses the
research situation on the probability of default model (PD) at home and abroad.
Then introduce the situation of internal rating in Hunan ICBC, analyses its
achievements and inadequacies in first and second phase of internal rating project.
Finally, we established corporate credit risk assessment model using logistic
regression model to empirically analysis the Hunan ICBC’s corporate customers. At
the time of model establish, we randomly selected 72 companies as modeling
sample group, 58 companies as test sample group. Though the processing of sample
group’s 2012 data, we got 40 financial indicators. The indicators were screened by
normality test, hypothesis test, correlation test using SPSS software, and regression
analysis were carried out by logistic regression model.
Though the processing of test sample group to verify the model,we find that
the accuracy rate is 84.5%. The corporate credit evaluation model established in
this paper is very effective for Hunan ICBC.
Key words: Hunan ICBC; Internal Credit Rating; Logistic Regression Model;
Credit Risk