Author(s): Shenggang Li

Originally published on Towards AI.

Clustered contextual bandits and tabular Q-learning with off-policy evaluation on real-world retail logs

Traditional market basket analysis (MBA) explains what tends to co-occur, but it does not decide what to do next. This paper advances from static rules to policy learning — we model shopping as a sequential decision problem and learn which product to recommend next to maximize business value (e.g., margin or revenue).

This article explores the application of reinforcement learning (RL) methods to enhance market basket recommendations by transforming traditional analysis into a dynamic, action-oriented framework. By implementing clustered contextual bandits and tabular Q-learning, the authors demonstrate how segments and contexts can be utilized to improve the recommendation process, ultimately aiming to maximize business metrics such as margin and revenue. Their findings indicate that the RL approaches outperform standard methods, providing significant uplift in recommendations based on a pipeline validated against real-world retail logs.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI