xgboost
xgboost copied to clipboard
Preprocessing within xgb.cv
I think in many cases some tasks like feature selection have to be done on each fold before training process. I guess in Python implementation is fpreproc option for passing function for data preprocessing (I haven't used it though). Can an option to perform preprocessing inside xgb.cv be added in later version?
A pull request is welcome.
Will converting the dmatrix into pandas dataframe inside fpreproc be an inefficient way to do this?
What's the current recommended way?