Personalized Point-of-Interest Recommendation Based on Social and Geographical Influence

With the rapid development of location-based social networks (LBSNs), personalized Point-of-Interest (POI) recommendation has become an important personalized service to help users explore the surrounding environment. To better solve the data-sparse problem of POI recommendation, the main idea of existing research is to use neural networks to fuse context information such as social relationships and geographical influence. However, the existing models are still inadequate in integrating context information, and few studies consider privacy protection against users' activity trajectories. To solve these problems, this paper proposes a POI recommendation algorithm, SGGCN, which integrates social relationships and geographical influence. Based on desensitization of user activity trajectory, this method uses a graph convolutional neural network to explicitly learn the collaborative signal between users and users, POIs and POIs, and users and POIs to alleviate the data-sparse problem. Experiments on two real data sets show a 10% improvement over state-of-the-art POI recommendation methods.