Research on Online Product Recommendation Algorithm Based on Graph Neural Network

In this paper, we propose a novel product recommendation algorithm leveraging a graph neural network combined with a self-attention mechanism to address the challenges of complex user behaviors, diverse product ranges, and strong inter-product correlations in online shopping platforms. This approach capitalizes on the GCN's capability to deeply model complex interactions by constructing a heterogeneous graph that maps the intricate relationships between users and products, where nodes represent users and products respectively. Integrating the self-attention mechanism with the graph convolutional network (GCN) technique allows for enhanced feature extraction of product data. The self-attention mechanism's advantages include its ability to adaptively focus on the most relevant nodes (i.e., users and products) for a given task, thereby capturing subtle connections between products effectively. It also addresses long-range dependencies, crucial for understanding users' long-term behavior patterns, and provides a flexible method for integrating information between nodes, enhancing the model's generalization capability and interpretability. In this work, the integration of GCN with self-attention is the primary distinction between existing graph networks. Such a model structure is more suitable for product recommendation tasks and enables the proposed model to outperform existing methods in product recommendation performance. Our experimental results indicate that this combined GCN and self-attention model outperforms traditional recommendation algorithms in capturing individual user needs and the nuanced relationships between products, significantly improving recommendation accuracy and coverage. This leads to an optimized user experience and higher conversion rates, demonstrating excellent performance and scalability in real-world business settings.