0
$\begingroup$

I had an idea to combine oversampling and undersampling in the following way:

Compute the average number of individuals in each class. For classes with a number of individuals greater than this average, undersample the class so that the number of individuals in this class equals the average. For classes with a number of individuals lower than this average, oversample the class so that the number of individuals in this class equals the average. Overall, all classes then have a number of individuals equal to the original average.

However, I cannot see any literature on this idea. Does this idea exist and does it have a name so that I may read about it?

$\endgroup$
0
$\begingroup$

It is called resampling and is quite famous for dealing with imbalanced datasets. See, for example, https://www.kaggle.com/rafjaa/resampling-strategies-for-imbalanced-datasets for implementation.

Congrats on good intuition!

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.