| Recommender systems face several challenges despite the significant technological advances in the wake of the booming digital media services. Issues like the strict cold start problem and sparse data are still noteworthy challenges in the field. Strict cold-start problems occur when there is no training and testing data available for new users and new movies. On the other hand, suggesting a relevant movie to an existing user is challenging when there is little interaction data available, leading to the sparsity problem. To mitigate these challenges, we propose a Multi-View Graph Neural Network (M-VGNN) model that incorporates both the attribute graph and interaction graph as input graphs, generates four embeddings from these graphs and uses them in a multi-view approach. Empirical results on benchmark movie datasets, i.e., ML-100K and ML-1M, show that our model yields improvements of 0.61% and 0.35% on RMSE in Item Cold Start and Warm Start scenarios for ML-100K dataset and 4.7% and 0.2% on RMSE in Item Cold Start and Warm Start scenarios for ML-1M dataset as compared to second-best model. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.