Use Borderline SMOTE to deal with the problem of sample imbalance

[onesmorewhite]

Why only two of these categories have been sample balanced？

from imblearn.over_sampling import BorderlineSMOTE
nn_k = NearestNeighbors(n_neighbors=29)
bsmote = BorderlineSMOTE(random_state=42,k_neighbors=nn_k)
X_train_smote,y_train_smote = bsmote.fit_resample(X_train,y_train)
print(sorted(Counter(y_train_smote).items()))
result：[(0, 136), (1, 136), (2, 3)]