A few months ago, I set myself the challenge of developing a new and intriguing algorithm in under 5 days. The outcome was GenBalance. The idea is shared by this paper -- aiming to preprocess a dataset to address both class imbalance and fairness.
I used the same overall methodology of GenFair, from which I (dirtily) recycled most of the code. Instead of removing DN and PP, we only increase DP and PN to balance both fairness and class distribution.
Alas, while it does technically work, the results fall short compared to the algorithm of Dablain et al. It's unlikely I'll revisit this project. Still, it was a fun ride!