Knowledge MaxxingResampling
Resampling addresses class imbalance by adjusting the training data itself, either duplicating minority-class examples or removing majority-class examples, to better balance the classes.
Prerequisite: Class imbalance
What matters
Because it changes what the model actually sees during training, resampling shifts the model's attention toward the minority class without needing to change the model or metric.
Example
Oversampling might duplicate rare fraud examples several times in the training set so the model sees them as often as legitimate transactions.
Common misconception
Resampling training data does not change the true real-world proportion of each class; it only changes what the model sees, so evaluation should still reflect real-world proportions.
Recall question
Why should evaluation still reflect real-world class proportions even after resampling training data?