I wouldn't drop any of the features when using a tree-based model, as a tree would have a hard time identifying the members of the class that is dropped.
Nice post Avi. That why I see it's a common practice now a days to just drop a column after implementing one hot encoding.
But I still feel one hot encoding creates lot of sparse columns.
Also I was looking at your paid subscription newletter. It would be good if you made plan for Indian audience as well, and price it accordingly. Keep up the good work
I wouldn't drop any of the features when using a tree-based model, as a tree would have a hard time identifying the members of the class that is dropped.
Nice post Avi. That why I see it's a common practice now a days to just drop a column after implementing one hot encoding.
But I still feel one hot encoding creates lot of sparse columns.
Also I was looking at your paid subscription newletter. It would be good if you made plan for Indian audience as well, and price it accordingly. Keep up the good work