Identifying Complete Individual Trajectories Using Multi-day Cellular Network Data

Abstract

Abstract. Individual Trajectory is the foundation of traveling behavior analysis. Cellular network data contains sufficient spatio-temporal information, which is widely used for trajectory analysis nowadays. However, we found in practice that most of the users' records are incomplete and fragmented. This is because mobile phones leave records of their connected cell tower only when some specific event occurs, such as making calls or sending messages. This paper present a method for identifying complete individual trajectories using fragmented trajectories extracted from multi-day cellular network data. This method firstly extracted a user's all fragmented trajectories when he/she move from one place to another from multi-day data. Secondly, we proposed a greedy algorithm to joint all the fragmented trajectories into a complete trajectory. Eventually, we adopted a modified Kalman filter algorithm to smooth the trajectory. In the end, we experimented with real data to validate our approach. The results show that our approach improves the useable data volume from 17% to 63% and reduce 36% of the trajectories' average bias distance.