摘要
以商业银行贷款数据为基础,采用聚类分析、多元判别和Logistic回归方法构建了我国商业银行的信用风险评级模型。实证检验表明这三种方法都能较准确地对商业银行的信用风险评级资料进行预测。对正常、关注、次级、可疑、损失五类不同样本的总体预测精度分别为83.4%、72.05%和68.14%。三种方法对五类不同样本的预测存在相同的趋势,即对正常类和损失类样本的预测准确率较高,对中间三类样本的预测准确率较低。
Based on the adequate loan data from X commercial bank, this article build the credit risk evaluation model for China' s commercial banks by adopting clustering analysis, discrimination analysis and logistic regression. The empirical test shows that all the three approaches can conduct accurate forecast of the credit risk evaluation data from commercial banks. The general forecast accuracy of three approaches for the five samples (i.e. normal, concerned, secondary, doubtful, and loss) are respectively 83.4%, 72.05%, and 68.14%. There is the same prediction tendency of the three approaches for the five samples: a relatively high accuracy for normal and loss samples but relatively low accuracy for the three samples in the middle.
出处
《河北经贸大学学报》
2005年第4期41-45,共5页
Journal of Hebei University of Economics and Business
关键词
信用风险
信用评级
逻辑回归
判别分析
聚类分析
credit risk
risk rating
logistic regression
discrimination analysis
clustering analysis
