How would you handle a dataset where the target class (for example, fraudulent activity) represents less than 1% of the total observations?
