This document discusses a model for predicting service ratings using location-based social networks (LBSNs) that leverage mobile users' geographical data. It identifies three key factors affecting rating behavior: user-item geographical connections, user-user geographical connections, and interpersonal interest similarity, and integrates them into a unified prediction model. Experimental results on the Yelp dataset show the proposed model outperforms existing recommendation systems, particularly addressing the cold start and sparsity issues.
Related topics: