luorongjin 发表于 2011-12-8 17:22 
你好！你的回答很好，学习了！你能简单说明下logistic 生存模型的方法和运用吗？谢谢！
For example, the popular FICO score is built on a logistic model. In this model the dependent variable (event) is defined as 90+ past due or worse. It has 24 month aging or history. The independent variables are in five catagries and around 20 variable or measures,
1) credit history --- usually measured by oldest tradeline. The longer the better.
2) Payment history ---- measured by number of your payment made on time, later, etc.
3) Credit utilization ---- measured by revolving debt in a ratio or detb to credit limit.
4) Credit inquire --- The more time one search for credit, the less score will be.
5) Types of credit used --- if one can manage different type of credit(mortgage, auto loan etc.) well, he will have better/higher score.
All data are from credit bureaus.
Other bigger banks may have developped their own score models. The dependent variable(event) may be defined differently and information set is also different. For example, a mortgage company may have income, appraisal report, credit bureaus (fico score) information, employment letter and tax return form. But the methodology is the same.
Usually a mortgage loan takes 15 ~ 30 years to mature. The survival analysis is needed to take additional score adjustment.
The macro economic situation plays a very important role in a loan performance. For example umemployment rate goes higher, you will see defaults goes high. Then we need to quantify its impacts.
扫码加好友，拉您进群
全部版块
我的主页
收藏
