This adaptation allows the user to pass categorical_features to which the SMOTE-NC procedure proposed in the original SMOTE paper was applied categorical features. It works in a similar fashion as imblearn.over_sampling.SMOTENC for categorical features. It is passing both existing tests and the ones originally used to test imblearn's SMOTENC.
BTW, the new tests include sparse matrix inputs, I adapted a few parts of the code to have it support sparse inputs (as referenced in issue #1 ). Honestly I'm not entirely sure how useful this is, very rarely have I used sparse matrix formats, but it's there anyway.
This adaptation allows the user to pass
categorical_featuresto which the SMOTE-NC procedure proposed in the original SMOTE paper was applied categorical features. It works in a similar fashion asimblearn.over_sampling.SMOTENCfor categorical features. It is passing both existing tests and the ones originally used to test imblearn's SMOTENC.BTW, the new tests include sparse matrix inputs, I adapted a few parts of the code to have it support sparse inputs (as referenced in issue #1 ). Honestly I'm not entirely sure how useful this is, very rarely have I used sparse matrix formats, but it's there anyway.