GraphFusionSBR: Denoising Multi-Channel Graphs for Session-Based Recommendation

2026-01-01

Expert systems with applications

AI Summary

This work addresses the challenge of inaccurate user intent modeling in conversational recommendation systems, which stems from sparse interactions and noisy signals. To this end, the authors propose a multi-channel graph fusion model that jointly leverages knowledge graphs, session hypergraphs, and session line graphs. The approach enables cross-channel collaborative representation learning through adaptive edge pruning, intra-session attention-based denoising, and a mutual information maximization mechanism between hypergraph and line graph structures. Evaluated on both e-commerce and multimedia recommendation scenarios, the method significantly improves recommendation accuracy. Its core innovations lie in the integration of heterogeneous graph structures and an adaptive denoising strategy. The implementation has been open-sourced to facilitate reproducibility.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Graph-based Machine Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Problem

• session-based recommendation
• item interaction dominance
• noisy sessions
• graph denoising
• multi-channel graphs

Innovation

• multi-channel graph
• adaptive denoising
• session-based recommendation
• hypergraph
• mutual information maximization