04-04-2015 07:23 AM
I am in the process of developing a predictive model. I need your help understanding rolling performance window. The objective of the model is to identify customer attrition in retail segment. Observation period is 12 months ( March 2013 to Feb 2014). Fixed Performance Window is 6 months (April 2014 to Sep 2014). While exploring attrition data, there is some seasonality in average attrition numbers. I was asked to take rolling performance window. Does it mean i should take 6 performance windows and develop 6 models? How would i come up with final model out of 6 models? How rolling performance window works? How it helps improving model robustness?
I understand it's not a SAS question. I posted this question here knowing it's a large analytics community.
Thanks in anticipation!