Location Recommendations Based on Multi-view Learning and Attention-Enhanced Graph Networks

Personalized location recommendation plays a very important role in location-based social networks, from which both users and service providers can benefit from. In spite of the significant endeavors that have been made towards acquiring knowledge on location attributes and user inclinations, it is still faced with serious data sparsity problems. In this paper, we propose a personalized location recommendation model based on graph neural networks with multi-view learning to obtain effective representations of mobile users and locations from different heterogeneous graphs. We also design an attention-enhanced mechanism to explore the implicit interactions between mobile users and locations themselves. Conducting adequate comparative experiments on two real-world telecom datasets has demonstrated that our model achieves superior performance. Additionally, our model has been proven effective in addressing data sparsity issues.