Interest Aware Dual-Channel Graph Contrastive Learning for Session-Based Recommendation

The key issue of session-based recommendation (SBR) is how to efficiently predict the next interaction item based on the item sequence of anonymous users. In order to mine the complex multivariate relationship between items and sessions, we propose a novel model for session-based recommendation named Interest aware Dual-channel Graph Contrastive learning (IDGC). By generating hypergraph and global graph, we focus on item relationships in different aspects, and we create the dual-channel interest-item embedding learning module to dig the higher-order relationships between items and users’ interests. To deal with the problem of long-distance information transmission between non-adjacent items, we set the interest node in each session for interest awareness and base on the contrastive learning strategy to enrich the information of the two graphs. At the same time, we exploit position information and time interval information to enhance the session representation. Extensive experiments show that IDGC has significant performance improvement on all evaluation metrics on three benchmark datasets.