Location Recommendation Incorporating Temporal and Spatial Effects

Services using the position information of users have been popularized by the spread of smart phones and tablet devices. One such service is location based social networking (LBSN). As users of LBSN have increased, vast amounts of check-in data have become available that record where users have been and when. In this paper, we assume that a user's check-in behavior is influenced by temporal and geographical factors. For the temporal factor, we consider the temporal impact to be different for each location and understand it by smoothing check-in data on the basis of Gaussian distribution. On the other hand, for the geographical factor, we assume that a user's check-in behavior depends on the area he visits on a daily basis and propose methods to reflect detected user's behavior range in the conventional system by utilizing spatial effects. In addition, we demonstrate that integrating these factors further improves recommendation accuracy. In our experiments, we show that our proposals outperform the state-of-the-art location recommender systems by using two LBSN datasets.