I'm training an XGBoost on a dataset of 2 millions of rows. But it's too bad to predict classes with low frequency. Is there any parameter on R which can help me to resolve that problem?
Asked
Active
Viewed 24 times
0
-
You need to have a balanced data set. Try oversampling or undersampling tehniques – DollarAkshay Jan 04 '19 at 09:20
-
@AkshayLAradhya I have already done that but the fact is that I have a class of 2% in the dataset. HoW can I handle prediction with this class? – bruuh Jan 12 '19 at 10:24