Extracting typical users' moving patterns using deep learning

When GPS devices are widely integrated into smart phones, researchers stand a big chance of collecting massive location information, that is necessary in studying users' moving behavior and predicting the next location of the users. Once the next location of a user can be determined, it can serve as input for many applications, such as location based service, scheduling users access in a mobile network or even home automation. One important task in predicting the next location is to identify typical users' moving patterns. In this paper, we propose a novel method to extract the patterns using deep learning. Experiment results show significant performance improvement of the proposed method compared to the classical principal component analysis method.