Extract feature weights from classifier model



Is there a way to extract the relative weights of each feature used for classification, preferably from random forest?


You can do this for FastGentleBoosting, but none of the others at the moment (though I talked to the current maintainer and she believes it’ll be possible to add). I’ve opened an issue requesting the feature here, if and when the functionality is added that will be the place it’s reported. Sorry!


Sounds great - Thanks! I’ll see if I can get similar classification out of FastGentleBoosting.