Enhanced Session-Based Recommendation with Global Context and Dynamic Memory

Session-based recommendation systems predict what a user might be interested in next based on their current behavior. However, many existing methods face issues, such as insufficient use of overall context and difficulty in extracting key features due to noise interference. To address these challenges, this paper introduces a model called GC-DMRec. This model combines global context with gated slot attention mechanisms, effectively integrating local session graphs with global context graphs to track item changes both within and across sessions. By using gated slot attention, GC-DMRec focuses on critical time points, extracting more representative features while reducing noise. Experimental results on datasets such as Diginetica, Tmall, and Nowplaying demonstrate that GC-DMRec outperforms other existing methods in terms of accuracy and reliability.