GNNRec: a graph-based neural network model for efficient and robust product recommendation

Abstract

Product recommendation systems are fundamental to modern e-commerce platforms, yet existing approaches face significant challenges related to data sparsity, cold-start problems, and scalability limitations in large-scale applications. To address these issues, this paper introduces GNNRec, a graph-based neural network model for efficient and robust product recommendation. GNNRec is a comprehensive recommendation framework that extends the LightGCN architecture by incorporating multi-relational user–item interactions, heterogeneous node embeddings, and multi-task learning. The model integrates rich semantic features and edge-weighted graph propagation to capture complex user preferences and item characteristics effectively. GNNRec mitigates cold-start effects through graph-based propagation and content-aware initialization and addresses data sparsity via adaptive sampling strategies. Extensive experiments on large-scale real-world datasets demonstrate that GNNRec significantly outperforms state-of-the-art baselines, achieving an AUC of 0.916 and consistently superior results across ranking metrics. In particular, GNNRec attains NDCG@10 of 0.0428, corresponding to a 4.4% improvement over the strongest baseline (NGCF at 0.0410). Moreover, GNNRec exhibits strong robustness under data sparsity (99.988%), improved cold-start performance (16.3% gain for users with 5–10 interactions and viable recommendations for zero-interaction users), and temporal variations, while achieving 42.4% faster training and 33.3% lower memory usage, confirming its scalability and suitability for practical deployment.