Looking for binary class datasets with high class imbalance, that also have intra-class imbalance in the minority class

Question

Newbie question alert...

For a college project I want to compare a few variants of SMOTE in terms of how much they improve classification of the minority class, over using random oversampling.

I have a specific interest in the idea that the minority class may contain small disjuncts that may themselves exhibit imbalance within the class.

I am already looking at the credit card fraud dataset on Kaggle (https://www.kaggle.com/mlg-ulb/creditcardfraud)

Can anyone please point me towards other datasets that have the following kinds of properties:

• a reasonably large number of examples (ideally at least a few thousand)
• have only two class labels
• are highly imbalanced, i.e. the minority class is severely under-represented
• ideally the minority examples would have some intra-class imbalance too

Or even better, is there any kind of good search tool out there for finding datasets based on these kinds of characteristics?

Answer 1

The imblearn.datasets package (documentation is here) has a function called fetch_datasets() which is described as:

fetch_datasets allows to fetch 27 datasets which are imbalanced and binarized

I do not know them in sufficient depth to know whether they meet all four of the criteria you've listed, but these may be a good first place to start with.