Logical error in selecting the most frequent categorical features based on the most frequent nearest neighbours of the majority class.

[RaghuSpaceRajan]

imbalanced-learn/imblearn/over_sampling/_smote.py

Line 1039 in f17107e

col_sel = rng.choice(col_max == col_max.max())

Doing a choice() on col_max == col_max.max() returns a True or False and not the index of the wanted column.
I replaced it with:
col_sel = rng.choice(np.where(col_max == col_max.max())[0])