首页|Comparison of Corporate Default Probability Modeling Methods
Comparison of Corporate Default Probability Modeling Methods
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
In banks’ internal-rating based approach to measure credit risk， modelers mainly use logistic regression and discriminant analysis for PD modeling。 We compare the required assumptions and structural differences between the two statistical methodologies。 We suggest transformations when desired conditions are not satisfied。 Further discussions are made on other modeling methods， including optimizing on Information Value (i。e。 relative entropy) as an objective function， CART， and neural network。 To demonstrate and validate our results， we select two sets of corporate data: one from a local bank’s small and mid-sized companies loans， and the other from manufacture sector of public companies。 Finally we briefly discuss the impact of different modeling approaches on advanced risk management methods using PD as a building block。
probability of defaultpredictive modelingdiscrimant analysislogistic regression
Qun XIE
展开 >
School of Economics and Management, Tsinghua University, Beijing, 100084, China
International Conference of Management Science & Applications;ICMSA 2005
NETL
