Xgboost node splits on a value that out of feature range? -
i have features range 0 1. when dump model, find nodes split features using "feature < 2.00001". xgboost scale feature or add value feature? or why 2.00001 chosen split? thanks~
xgboost has separate splits based on values of feature , whether or not feature missing.
this when xgboost wants split on whether or not feature missing, , not based on value.
Comments
Post a Comment