Given a highly imbalanced dataset, how would you structure your training and validation sets in Python?
