I have 4 categorical features, one of which is ordinal. I think SMOTENC uses one-hot encoder for all features passed to categorial_features(). How do I deal with ordinal categories?
I'm currently working with this format:
from imblearn.over_sampling import SMOTENC
X_resampled, y_resampled = SMOTENC(categorical_features = ['nominalcat1', 'nominalcat2', 'nominalcat3']).fit_resample(X_train, y_train)