Exploiting Item Relationships with Dual-Channel Attention Networks for Session-Based Recommendation.

Session-based recommendation (SBR) is the task of recommending the next item for users based on their short-term behavior sequences. Most of the current SBR methods model the transition patterns of items based on graph neural networks (GNNs) because of their ability to capture complex transition patterns. However, GNN-based SBR models neglect the global co-occurrence relationship among items and lack the ability to accurately model user intent due to limited evidence in sessions. In this paper, we propose a new SBR model based on Dual-channel Graph Representation Learning (called DCGRL), which well models user intent by capturing item relationships within and beyond sessions respectively. Specifically, we design a local-level hypergraph attention network to model multi-grained item transition relationships within a session by using sliding windows of different sizes. The experiments demonstrate the effectiveness and the efficiency of our proposed method compared with several state-of-the-art methods in terms of HR@20 and MRR@20.