Categorical features - Githubissues

Taakn commented 6 years ago

Hello,

More of a question than an issue, but how do you handle categorical features?

Also seems like y needs to be an int? Otherwise if I leave it as float I am getting an error.

Thanks!

SymbolicSquared commented 5 years ago

For categorical features you will need to vectorize/dummy code, just like you would if you plan to use any classifier in sklearn. I suggest pandas get_dummies method.

ThomasBury commented 4 years ago

Hi,

regarding categorical:

large cardinality biases impurity based var.imp --> would it be possible to use permutation imp instead (available in scikit)?
dummifying is not advised for tree-based models (making the trees unstable and deeper than expected) see https://roamanalytics.com/2016/10/28/are-categorical-variables-getting-lost-in-your-random-forests/. Using lightgbm, you can encode as integer and use the native method to handle cat. pred. Alternative: use https://github.com/scikit-learn-contrib/category_encoders

danielhomola commented 4 years ago

Submit a PR with tests and I'll have a look.

scikit-learn-contrib / boruta_py

Categorical features #41