Where will I go next?: Predicting future categorical check-ins in Location Based Social Networks

Models to predict the future location of users have been developed in the past few decades. However, these efforts cannot drive applications related to location-based targeting since they focus on flat geographic prediction with no semantic information. With the emergence of Location Based Social Networks (LBSN) geographical data can be supplemented with contextual information. An efficient location predictor might bring numerous opportunities and commercial benefits. In this work we propose two simple predictors modeling future geo-contextual user behavior. The algorithms have two outputs: first the most likely next visit in terms of category and second the expected time frame, in when, such a visit may occur. The predictors use categorized user activities as unique checkins at specific times. Using real data obtained from the commercial LBSN (FourSquare), we show the efficiency of the algorithms.