Behavioral Credit Scoring - Default Yearmonth

[dincooo]

Hi all,

We are developing a behavioral credit scorecard and we are not sure about including the data (Default_Flag = 1) for the same customer for different yearmonths.

For example, tihs customer defaults on 201505 and I’m including the same customer to my dataset with default flag =1 for five times.

Is it true or do I have to choose one default_flag =1 observation randomly?

Customer_ID	Year_Month	Default_Flag
101	201501	0
101	201502	0
101	201503	0
101	201504	0
101	201505	1
101	201506	1
101	201507	1
101	201508	1
101	201509	1

Any ideas would be appreciated 🙂

Many thanks,

Onur

[SASKiwi]

It looks like you are tracking a single customer over time. In that case you should flag the first month that they go into default as default_flag = 1 and it should stay like that for all following months until the customer recovers (no longer in default) or the customer's loan is written off as a bad debt and closed. In the latter case you would remove the customer entirely from your data at this point as they would no longer have any lending.