[python] ignore pandas ordered categorical columns by default #2115
Conversation
| else: | ||
| categorical_feature = list(categorical_feature) + list(cat_cols) | ||
| categorical_feature = list(categorical_feature) + cat_cols_not_ordered |
There was a problem hiding this comment.
I wonder whether this is a good decision to add default categorical columns even in case user explicitly specifies categorical features. Firstly, this feature is not documented. Secondly, it seems that this approach makes some discomfort for users: the only way to disable default cat features is to convert them in source dataframe. I think it's more comfortable just not include them in list of cat features. The intuition behind this is that user want either use default cat features in cat columns, or specify cat features in param, but not both at the same time.
There was a problem hiding this comment.
good point, this indeed is not comfortable.
There was a problem hiding this comment.
@guolinke Cool! I'll address this in next PR, OK?
Closed #2108.