0
$\begingroup$

I have a dataset of 20 features. I have optimised Random Forest and the parameters it uses are:

  • number of trees: 167

  • max # of features used to split a node: 9

If I can find out what features where used to split a node and if these same 9 feaures are used in all the trees; can I skip a feature selection method? I mean I just eliminate the 11 features not used.

$\endgroup$
  • 1
    $\begingroup$ Try to build multiple decision trees and plot them to see the variables... $\endgroup$ – user9292 Jan 4 '17 at 4:25
3
$\begingroup$

Random forest chooses a random sample of the features to split on, picking the best split among the sample. It's plausible that all 20 features are used by the model across all of the trees.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.