Exploiting Spatiotemporal Features to Infer Friendship in Location-Based Social Networks

The popularity of smart phone has brought the pervasiveness of location-based social networks. A large number of check-in data provides an opportunity for researchers to infer social ties between users. In this paper, we focus on three problems: (1) how to exploit fine-grained temporal features to characterize people's lifestyle. (2) how to use weekday and weekend check-ins data. (3) how to effectively measure the fine-grained location weight. To tackle these problems, we propose a unified framework STIF to infer friendship. Extensive experiments on two real-world location-based datasets show that our proposed STIF framework can significantly outperform the state-of-art methods.