Abstract:Sequential recommendation aims at understanding user preference by capturing successive behavior correlations, which are usually represented as the item purchasing sequences based on their past interactions. Existing efforts generally predict the next item via modeling the sequential patterns. Despite effectiveness, there exist two natural deficiencies: (i) user preference is dynamic in nature, and the evolution of collaborative signals is often ignored; and (ii) the observed interactions are often irregularly-sampled, while existing methods model item transitions assuming uniform intervals. Thus, how to effectively model and predict the underlying dynamics for user preference becomes a critical research problem. To tackle the above challenges, in this paper, we focus on continuous-time sequential recommendation and propose a principled graph ordinary differential equation framework named GDERec. Technically, GDERec is characterized by an autoregressive graph ordinary differential equation consisting of two components, which are parameterized by two tailored graph neural networks (GNNs) respectively to capture user preference from the perspective of hybrid dynamical systems. On the one hand, we introduce a novel ordinary differential equation based GNN to implicitly model the temporal evolution of the user-item interaction graph. On the other hand, an attention-based GNN is proposed to explicitly incorporate collaborative attention to interaction signals when the interaction graph evolves over time. The two customized GNNs are trained alternately in an autoregressive manner to track the evolution of the underlying system from irregular observations, and thus learn effective representations of users and items beneficial to the sequential recommendation. Extensive experiments on five benchmark datasets demonstrate the superiority of our model over various state-of-the-art recommendation methods

Learning the Dynamics in Sequential Recommendation by Exploiting Real-time Information

Exploiting Structural and Temporal Influence for Dynamic Social-Aware Recommendation

Learning Dual Dynamic Representations on Time-Sliced User-Item Interaction Graphs for Sequential Recommendation

Learn over Past, Evolve for Future: Search-based Time-aware Recommendation with Sequential Behavior Data

Time Matters: Sequential Recommendation with Complex Temporal Information

Long Short-Term Preference Modeling for Continuous-Time Sequential Recommendation

DSIM: Dynamic and Static Interest Mining for Sequential Recommendation

HTP: Exploiting Holistic Temporal Patterns for Sequential Recommendation

Modeling Dynamic User Preference via Dictionary Learning for Sequential Recommendation

Dynamic Item-Based Recommendation Algorithm with Time Decay

Learning Heterogeneous Temporal Patterns of User Preference for Timely Recommendation

Attention-based dynamic user modeling and Deep Collaborative filtering recommendation

Adaptive time series prediction and recommendation

Learning Graph ODE for Continuous-Time Sequential Recommendation

Multi-interest sequential recommendation with contrastive learning and temporal analysis

Déjà vu: A Contextualized Temporal Attention Mechanism for Sequential Recommendation

Multi-Behavior Sequential Recommendation with Temporal Graph Transformer

Dynamic Graph Neural Networks for Sequential Recommendation

A Generic Learning Framework for Sequential Recommendation with Distribution Shifts

Continuous-Time Sequential Recommendation with Temporal Graph Collaborative Transformer