Learning Location Correlation from GPS Trajectories

People’s location histories imply the location correlation that states the relations between geographical locations in the space of human behavior. With the correlation, we can enable many valuable services, such as location recommendation and sales promotion. In this paper, by taking into account a user’s travel experience (knowledge) and the sequentiality that locations have been visited, we learn the location correlation from a large number of user-generated GPS trajectories. Using the location correlation, we conduct a personalized location recommendation system, which is evaluated based on a real-world GPS dataset collected by 112 users over a period of 1.5 years. As a result, our method outperforms that using the Pearson correlation.