search

csinva / imodels

Interpretable ML package 🔍 for concise, transparent, and accurate predictive modeling (sklearn-compatible).
https://csinva.io/imodels
MIT License
1.39k stars 122 forks source link

support categorical input features? #77

Open Yannahhh opened 2 years ago

Yannahhh commented 2 years ago

It seems that this package does not support categorical variables? Right?

csinva commented 2 years ago

That is correct, we are working on this but for now if you want to use categorical variables you'll have to first one-hot encode them.

paularindam commented 2 years ago

I was trying BoostedRulesClassifier for categorical labels. It does not work even after one hot. I assume this only supports Binary Classification. One can use one hot but only for input features.

csinva commented 2 years ago

Good point! You are correct my above comment was about categorial feature variables which an be used with one-hot encoding.

Most of the models here only support binary classification (see this table) for details.

paularindam commented 2 years ago

Is there anyone which supports multi-class classification? I can see binary classification and regression. Do you plan to add multi-class support ?

csinva commented 2 years ago

We do plan to add it although it may not be for another month - RuleFit, SLIM, FIGS, and HSTree (among others) should easily support multi-class, but we haven't been testing it so there may be an issue when using them for the multi-class setting.

paularindam commented 2 years ago

Thanks for the prompt response.

mepland commented 1 year ago

@paularindam FIGS just received categorical support in https://github.com/csinva/imodels/pull/151