Recommending Popular Locations Based on Collected Trajectories

Data gathered from location-aware devices, such as GPS, create opportunities for researchers to extract interesting information from the movement of objects. Popular places are regions that are visited for long durations of time or by a large number of objects. For human trajectories, such popular places are of interest in location recommender systems. In this paper, a number of input trajectories are preprocessed to efficiently answer queries about popular places. Each query specifies one potential popular place and the minimum and maximum duration of a visit. The answer to any such query is the number of visits to the corresponding popular place. We present algorithms for this problem and experimentally evaluate them on real-world data sets. One advantage of the algorithms presented in this paper for location recommender systems is that, unlike most of them, it works even when social network databases are unavailable or unreliable.