[BUG] Oversampling warning spurious / misleading

[amueller]

Describe the bug

If I resample a dataset 10x for each class, I might get a warning like

After over-sampling, the number of samples in class 1 will be larger than the number of samples in the majority class

Which is true, but I don't know why it deserves a warning. Maybe the criterion for the warning should be that after sampling another class is bigger than the majority class? But not sure.

Expected Results

No warning.

Actual Results

Warning, potentially for each other class. If the classes are of similar sizes, I will likely get a warning for each class.

[chkoar]

I suppose that you are referring to the fact that since one could do it intentionally there is no point for warning in this stage.

[amueller]

Well, yes, but I also don't understand what the intuition of the warning would be? Like why is it a bad thing that in the new data some classes are larger than the majority in the old data? Like what does that tell us?

[glemaitre]

Like what does that tell us?

I assume that my thinking at the time was that you create so many sample that now you still have an imbalanced problem but up-side-down. I agree that we can remove the warning.