Variable selection is a very important task in designing a good model. Variable selection gives the modeler the ability to focus on the most important aspects of the data first and NOT be overwhelmed by the enormity of the data itself. However, there is no single algorithm or method that can guarantee us the most optimal set of variables. Thus, variable selection should be made based on many different approaches as opposed to a few.
This session will talk about various approaches to Variable Selection and ways to test if those approaches are giving you reasonable results. Numerous examples will be presented to make the session more meaningful to actuaries.


Moderator/Panelist:
Ravi Kumar, Milliman Consulting

Panelist:
Mark S. Richards, Director of Analytics, ISO Innovative Analytics