Exploring uplift modeling with high class imbalance

Abstract Uplift modeling refers to individual level causal inference. Existing research on the topic ignores one prevalent and important aspect: high class imbalance. For instance in online environments uplift is used optimally target ads discounts, but very few users ever end up clicking an ad or buying. One common approach deal with imbalance classification by undersampling dataset. In this work, we show how can be extended modeling. We propose four methods for compare proposed empirically when some have a tendency break down. key observation that accounting particularly random forests, which explains poor performance of model earlier works. Undersampling also crucial class-variable transformation based models.