-
Curious, does the DecisionTreeEncoder (see here) work univariate-ly as the doc seems to suggest (i.e. for every feature separately) or does it consider interactions as it should be? I hope this makes sense? Thanks. |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 2 replies
-
It works univariately. |
Beta Was this translation helpful? Give feedback.
-
Thanks for the reply. Will accept as answer. So this is with the expectation for the upstream model to squeeze more signal out of this? IMHO information could be lost though? I am more an R person and have used decision trees in the past and simply fed in the features and used the terminal numbers, which are ordinal w.r.t. the target for regression. In Python, it appears that DecisionTreeRegressor cannot deal with categorical/object types and e.g. one hot has to be used, which is a bit silly IMHO when one deals with very granular categorical features. Are you aware of decision trees (for regression) that can deal with categorical/object types in Python? Thanks. |
Beta Was this translation helpful? Give feedback.
It works univariately.