When to use oversampling or undersampling

1. Should I use undersampling or oversampling?
2. When or why should we use oversampling?
3. When should you do undersampling?
4. Is it a good idea to oversample?

Should I use undersampling or oversampling?

Oversampling methods duplicate or create new synthetic examples in the minority class, whereas undersampling methods delete or merge examples in the majority class. Both types of resampling can be effective when used in isolation, although can be more effective when both types of methods are used together.

When or why should we use oversampling?

There are three main reasons for performing oversampling: to improve anti-aliasing performance, to increase resolution and to reduce noise.

When should you do undersampling?

Undersampling is appropriate when there is plenty of data for an accurate analysis. The data scientist uses all of the rare events but reduces the number of abundant events to create two equally sized classes.

Is it a good idea to oversample?

Oversampling is a well-known way to potentially improve models trained on imbalanced data. But it's important to remember that oversampling incorrectly can lead to thinking a model will generalize better than it actually does.