EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

2834. Exploring Learning to Rank for Optimal Treatment Allocation

Invited abstract in session MC-31: Causal Machine Learning, stream Analytics.

Monday, 12:30-14:00
Room: 046 (building: 208)

Authors (first author is the speaker)

1. Toon Vanderschueren
Research Centre for Information Systems Engineering (LIRIS), KU Leuven
2. Wouter Verbeke
Faculty of Economics and Business, KU Leuven
3. Felipe Moraes
Booking.com
4. Hugo Proença
Booking.com

Abstract

Efficiently allocating treatments while accounting for operational constraints constitutes an important challenge across various domains. In marketing, e.g., using promotions to target potential customers and boost conversions is limited by the available budget. While much research focuses on estimating causal effects, there is limited work on learning how to optimally allocate treatments. Existing methods for uplift modeling or causal inference that estimate treatment effects do not consider how the estimated effects relate to decisions made based on these estimates. Therefore, a potential downside of these methods is that the resulting predictive model may not be aligned with the operational context, resulting in prediction errors being propagated to the optimization problem and, subsequently, a suboptimal allocation policy. We explore an alternative approach based on learning to rank. In this approach, the idea is to directly learn an allocation policy that prioritizing instances for treatment in terms of their treatment effect. We explore different ranking metalearners and propose an efficient sampling procedure for the optimization of the ranking model to scale our methodology to large-scale data sets. We validate our methodology empirically and show its effectiveness in practice through a series of experiments on both synthetic and real-world data.

Keywords

Status: accepted

Back to the list of papers