Suppose you are fitting a binary classifier to a data set. You discover that the outcome variable is severely imbalanced. Which one of the following is a valid way to deal with this imbalance? Use a resampling method on the training data only. Use a resampling method on the testing data only. Use a resampling method on the original dataset. Use a resampling method on both the training and testing data