View article

Prediction of future locations from traces of human mobility has significant implications for location-based services. Most existing research in this area focuses on predicting the next location or the destination rather than the entire route. This paper presents a temporal frequent-pattern tree (TFT) method for predicting future locations and routes. We evaluate the method using a real-world dataset containing location data from 50 users in a city. Our results show that for more than 91% of the users, the accumulated average distance between the actual and predicted locations is less than 1000 m (). The results also show that the model benefits from similarities between users’ movement patterns.