There is a module named SMOTE(Synthetic Minority Oversampling Technique ) which increases the number of samples of undersampled data, I guess we should choose a feature(a feature to be predicted) which is underrepresented. How to choose it? There seems to be no option on choosing the column.