Modeling User Mobility via User Psychological and Geographical Behaviors Towards Point of-Interest Recommendation

The pervasive employments of Location-based Social Network call for precise and personalized Point-of-Interest (POI) recommendation to predict which places the users prefer. Modeling user mobility, as an important component of understanding user preference, plays an essential role in POI recommendation. However, existing methods mainly model user mobility through analyzing the check-in data and formulating a distribution without considering why a user checks in at a specific place from psychological perspective. In this paper, we propose a POI recommendation algorithm modeling user mobility by considering check-in data and geographical information. Specifically, with check-in data, we propose a novel probabilistic latent factor model to formulate user psychological behavior from the perspective of utility theory, which could help reveal the inner information underlying the comparative choice behaviors of users. Geographical behavior of all the historical check-ins captured by a power law distribution is then combined with probabilistic latent factor model to form the POI recommendation algorithm. Extensive evaluation experiments conducted on two real-world datasets confirm the superiority of our approach over state-of-the-art methods.