Enterprise Miner - Interactive Grouping with many variables

[grezek]

While I've used EM/Scorecard to recreate scorecards (I'm in model validation), I've never had to select input variables from a large number of potential inputs.  I've tried Interactive Grouping determine the inputs with the highest IV's, but the process choked on the number of input variables (1,400).  What is the standard approach(es)?  In the data I can identify families of inputs.

Any suggestions are welcome.  Thank you.  -- George Rezek

[Ksharp · Accepted solution]

If I were you, I would use PROC HPGENSELECT or PROC PLS pickup the most 30-40 variables and Group these 30-40 variables by ScoreCard node.

[grezek]

Thank you very much. I'll investigate and let you know how it goes. Thanks for the help. -- George