摘要
本文通过将连续数值变量进行序别化转换赋值,并基于这些变量建立Log- it信用评分模型,通过使用统计量AUC值与条件熵比率来检验序别化转换前后所建立回归模型的违约预测力。结果发现,连续数值变量经序别化转换后可提高模型的违约预测力及其韧性。
This paper transforms continuous variables into their corresponding ordinal format firstly, then based on these ordinal variables, we use Logit regression to build credit scoring model. Thereafter, by AUC and CIER we have compared and measured accuracy of predicting default among models built on continuous variables and their ordinal transformation. The result shows, in credit scoring model, ordinalization can improve the ability of predicting default and enhance its robustness.
出处
《国际金融研究》
CSSCI
北大核心
2006年第8期60-65,共6页
Studies of International Finance
