UTBoost: A Tree-boosting based System for Uplift Modeling
Abstract: Uplift modeling is a suite of machine learning techniques that managers can use to predict the incremental impact of an action on a particular customer outcome. These models enable decision-makers to pinpoint which customer segments will most likely respond positively to a specific intervention, thereby enhancing the efficiency of resource distribution and boosting total gains. However, accurately estimating this incremental impact presents tangible difficulties since it requires assessing the difference between two mutually exclusive outcomes for each individual. In our research, we introduce two novel modifications to the established Gradient Boosting Decision Trees (GBDT) technique, which learn the causal effect in a sequential way and overcome the counter-factual nature. Both approaches innovate existing techniques in terms of ensemble learning method and learning objective, respectively. Experiments on large-scale datasets demonstrate the usefulness of the proposed methods, which often yielding remarkable improvements over baseline models.
Submission Length: Regular submission (no more than 12 pages of main content)
Assigned Action Editor: ~Fabio_Stella1
Submission Number: 2309
Loading