Sequential Recommendation Using Deep Reinforcement Learning and Multi-Head Attention

Raneem Sultan, Mervat Abu-Elkheir

Abstract: Recommender Systems have become a crucial part of many of our online interactions. From shopping for clothes, planning a trip, or deciding what to watch, recommender systems are aiming to help users navigate the overwhelming amount of options available online. The problem with most of the existing recommender systems is that they treat the recommendation process as a static one and make recommendations according to a fixed greedy strategy. This is a problem because user preferences are dynamic. In this paper, we aim to address this problem by modeling the recommendation problem as a Markov Decision Process (MDP) and solving it using deep reinforcement learning. Furthermore, we use multi-head attention to improve the recommendations. We conduct extensive experiments using the MovieLens real-world dataset and achieve an improvement of 6% over the state-of-the-art approach results in terms of precision@20.