NAH-GNN: A graph-based framework for multi-behavior and high-hop interaction recommendation.

Area of Science:

• Artificial Intelligence
• Computer Science
• Data Science

Background:

• Recommender systems are crucial for personalized marketing but struggle with complex user behaviors and sparse data.
• Traditional methods fail to capture diverse interaction types and higher-order dependencies effectively.

Purpose of the Study:

• To propose a novel recommendation model, MBH-GNN, to optimize personalized marketing strategies.
• To address limitations in handling complex user behaviors and sparse data in recommendation systems.

Main Methods:

• Constructing a multi-behavior interaction graph to integrate diverse user-item interactions (browsing, favoriting, purchasing).
• Employing neighborhood-aware modeling with dynamic behavior weighting for semantically rich embeddings.
• Incorporating high-hop relational learning to capture long-range user-item dependencies and contextual information.

Main Results:

• MBH-GNN significantly outperforms baseline methods on BeiBei and Tmall datasets, achieving HR@10 of 0.789 and NDCG@10 of 0.330 (BeiBei), and HR@10 of 0.773 and NDCG@10 of 0.319 (Tmall).
• Demonstrated exceptional robustness and adaptability in addressing data sparsity and cold-start scenarios.
• Achieved higher recommendation accuracy and diversity in complex scenarios.

Conclusions:

• MBH-GNN offers an efficient and scalable solution for personalized marketing.
• The model provides critical theoretical support and practical value for improving recommendation system performance and modeling complex user behavior.