04-24-2017 11:13 AM - edited 04-24-2017 11:13 AM
I am trying to build a data set to model customer's behavior after credit card application has been approved (existing customers).specifically I am trying to score existing customers so that when score falls below a certain threshold then we can reduce their line of credit.
The way it is explained In sas credit risk course is that I have to collect 3 years of data of the whole portfolio for example from 2013 to 2015 and then watch the behavior for year 1 in this case 2013 and then take a random sample in year 2 (2014) and lastly access performance (90+ days delinquent or <90) one year later from random sample in year 2 . for example if I have an account with 36 observations (3 years worth of data) I take one random sample in year 2 (2013) and access performance (90+ days delinquent or <90) one year later.
my question is how do I deal with accounts that don't have 36 observations. say a new account was booked in 2014 and has only 24 months worth of data, how do I deal with that?
Thanks in advance