UTBoost: Gradient Boosted Decision Trees for Uplift Modeling

TL;DR

This paper introduces UTBoost, a novel gradient boosting approach tailored for uplift modeling, which effectively estimates the incremental impact of actions on customer outcomes by addressing the counterfactual dilemma.

Contribution

It proposes two innovative modifications to GBDT that sequentially learn causal effects, improving uplift modeling accuracy over existing methods.

Findings

Consistent improvements over baseline models on large-scale datasets

Effective estimation of incremental impact in customer actions

Addresses the counterfactual dilemma in uplift modeling

Abstract

Uplift modeling comprises a collection of machine learning techniques designed for managers to predict the incremental impact of specific actions on customer outcomes. However, accurately estimating this incremental impact poses significant challenges due to the necessity of determining the difference between two mutually exclusive outcomes for each individual. In our study, we introduce two novel modifications to the established Gradient Boosting Decision Trees (GBDT) technique. These modifications sequentially learn the causal effect, addressing the counterfactual dilemma. Each modification innovates upon the existing technique in terms of the ensemble learning method and the learning objective, respectively. Experiments with large-scale datasets validate the effectiveness of our methods, consistently achieving substantial improvements over baseline models.