I am currently training a BDTG with k-fold cross-validation. After all the folds are trained, I am able to get an input variable ranking that uses ‘Separation’ as the metric:
: Ranking input variables (method unspecific)...
IdTransformation : Ranking result (top variable is best ranked)
: --------------------------------------------------
: Rank : Variable : Separation
: --------------------------------------------------
I have seen both in the TMVA Users Guide and in some other Root Forum examples (e.g. https://root-forum.cern.ch/t/the-rank-of-variables/28339
) that other metrics are available, one being ‘Variable Importance.’ How do we switch between different types of rankings? Is there some option to do this?
Thanks!
Thomas