Travel estimation using Control Signal Records in cellular networks and geographical information

We propose a method to estimate people travel modes and trajectories from control signal records (CSRs) generated at cellular base stations (BSs) when mobile phones send BSs any signal such as handover and data transmission. User locations can be roughly identified by BS IDs in the records as well as the locations of BSs. However, the location accuracy depends on BS cell size, which is not enough to estimate people travels. Therefore, our method leverages geographical information such as stations, railway networks and road networks. We firstly extract railway passengers and automobile travelers based on roughly estimated speeds and trajectories. Then, we further estimate trajectories for automobiles and boarded trains for railway passengers by applying maximum likelihood estimation. Simulation results show that our method achieves 88% recall and 89% precision for railway passenger identification and 85% accuracy for boarded train identification if each mobile phone's communication frequency exceeds 9 times per hour. For automobile travels, our method achieves 75% recall and 70% precision for identification and 64% accuracy for trajectory estimation.