Uplift Modeling from Separate Labels

TL;DR

This paper introduces a new uplift modeling method that estimates the impact of actions using only one type of label per instance, making it more practical for real-world applications.

Contribution

It proposes a novel uplift modeling approach that works with single-label data, addressing the challenge of obtaining joint labels in real-world scenarios.

Findings

The proposed estimator has a proven mean squared error bound.

Experimental results demonstrate the effectiveness of the method.

The approach broadens the applicability of uplift modeling in practical settings.

Abstract

Uplift modeling is aimed at estimating the incremental impact of an action on an individual's behavior, which is useful in various application domains such as targeted marketing (advertisement campaigns) and personalized medicine (medical treatments). Conventional methods of uplift modeling require every instance to be jointly equipped with two types of labels: the taken action and its outcome. However, obtaining two labels for each instance at the same time is difficult or expensive in many real-world problems. In this paper, we propose a novel method of uplift modeling that is applicable to a more practical setting where only one type of labels is available for each instance. We show a mean squared error bound for the proposed estimator and demonstrate its effectiveness through experiments.