SAS EM: using HP Forest / Random Forest for Variable Selection

[nelson_lee]

Dear all,

I would like to use Random Forest for Variable Selection, my datasets have over 600 variables

While when I run HP Forest Node (setting 50 as Max no. of Tree), it appear run time error as below. What should I do?

Error: Run time error was encountered. Please see the log in the node Results window for more details.
* ERROR: Run time error was encountered.  The system error returned was 1012.
* Please report unresolved problems to Technical Support.

One more question:

 - Are they any best practices for using HP Forest Node for Variable Selection? Thanks.

Best regards,

Nelson

[DougWielenga]

System error 1012 is a generic one that usually can be isolated to the data or variables.  Start by trying to run a regular Decision Tree and see if you encounter any issues.  If that runs, try and run a default HP Forest node.  If it fails, it would be helpful to see a model package created from the node that failed.  This is the first thing you are likely to be asked for by SAS Tech Support in this situation.  The model package can identify issues in any node leading up to the node where the problem was encountered.  

Follow the instructions to complete a model package from the End Groups node in SAS Note 46764 - Creating a model package (.SPK file) in SAS® Enterprise Miner(tm) which is available at 

http://support.sas.com/kb/46/764.html

Please note that the model package will be created even if an error message appears after completing the steps.  It will contain logs and settings for the nodes but it will not contain your raw data.   Attach the model package to your reply to allow further investigation. 

Hope this helps!

Doug