CSCL-GNN: Cross-View Semantic Contrastive Learning and Graph Neural Networks for Event Recommendation in Event-Based Social Networks

Event-based social networks (EBSNs) connect users through shared interests in real-world events. These platforms allow users to discover, join, and interact around events, but their complex relationships make accurate recommendations challenging. Traditional methods often modify network structures (e.g., randomly removing connections), risking information loss or noise injection, which harms model accuracy. Additionally, they ignore users with similar interests but weak social ties. To solve these issues, we propose cross-view semantic contrastive learning and graph neural networks for event recommendation (CSCL-GNN), a novel event recommendation model that combines graph learning with semantic analysis. First, we employ dual graph learners (LightGCN and multihead attention) to directly model the user–event interaction graph and the social relationship graph, avoiding the structural distortions caused by traditional data augmentation methods. Second, we leverage K-means clustering to identify latent user groups and integrate semantic information from both interaction and social views, reducing cross-view semantic discrepancies. Finally, our model achieves significant performance improvements (up to 2.1%) on five public datasets, with ablation studies demonstrating the effectiveness of each module.