Feature Selection for Regression

[Emanuele Olivetti]

I'm new to PyML and I'd like to use it for regression. How can I perform feature selection for regression problems? Is it available from featsel module? I reading the tutorial but classification seems to be the main topic. Could someone help me?

Thanks in advance,

Emanuele

[abh]

So far I haven't implemented feature selection for regression.
An RFE-like method should be easy to program -- it's possible that the current implementation would work as-is, but I haven't tried that.

-Asa

[Emanuele Olivetti]

It seems there is something to change. Trying to train a featureSelector using svm.SVR I got:
(ipythn logs follows)
---------------------------
/usr/lib/python2.4/site-packages/PyML/featsel.py in run(self, data, *options, **args)
    278     def run(self, data, *options, **args) :
    279
--> 280         if data.labels.numClasses != 2 :
    281             raise ValueError, 'RFE supports only two class problems'
    282        

AttributeError: 'Labels' object has no attribute 'numClasses'
----------------------------

Maybe some little change in the code is necessary. I'll try to spend some time on it but I'm a really newbie :)

Thanks anyway,

Emanuele