|
Say I have dataset with a numerical feature, that can only have integer values, for example, the number of links in a document. My goal is to classify the dataset, and I'm using a Random Forest for that. Does it make sense to bin that feature before feeding the dataset to the Random Forest to avoid over fitting? |
|
The RF does the binning internally when it builds the tree, but it's based on some defined criteria. Why not just let the RF handle that? |
