Thursday, April 19, 2007

Use of Economic and Environmental Variables in Modeling

To remove the seasonal effect, sometimes it becomes important to include seasonal variables into the modeling process. Here is a way I could think of for using such variables.

1. Handling of variables not related to credit worthiness or behavior of customers like environmental, economical or seasonal variables. This can be done in following two ways :

a. Building scoring model with all variables available but removing these variables from scoring. This way the bias in the sample due to these factors is removed and also these variables do not influence the score.
b. Building a logistic model with such variables only and take the residual DV. Build scoring model on the Residual DV with other variables.

Inclusion of these variables in the scoring process give different score for a individual or individuals having similar credit behavior coming at different season which may not be expected. We can reduce or increase cutoff in different season, if needed, to tackle such issues.

No comments:

Page Views from May 2007


© New Blogger Templates | Webtalks