RuleFit- huge error caused by user or incorrect code?

TMVA
1

Does anyone know if the code for RuleFit method has been fully debugged?
Has anyone used RuleFit as a classification scheme and if so have they gotten error rates comparable to the other classification methods?

From reading the code it looks like RuleFit is not fully polished and there are some errors in which error measures are returned so I want to know if the code does not work correctly or if I’m calling it incorrectly in my script. TMVA decision tree methods such as BDT and BDTG give error rates of 1.5-3% which is similar to the error that I get with other classification software on this data set. RuleFit however returns huge error rates of ~20%.

Is the error in the code or in my use of the code?

Thanks.

2

A few people have viewed this post but no one has offered a comment. I gather this means that RuleFit has not been used to any significant extent. Is this a correct conclusion?

3

You can try to post your message also to tmva-users@lists.sourceforge.net

Lorenzo

4

Thanks for the suggestion. I have sent out an email and hope that source will be more vocal. Thank you for your help.