Efficient and accurate fine grained trip identification for metro passengers from cellular trajectory data

Gaining a comprehensive understanding of individuals' fine-grained metro trips within a city is crucial for effective urban planning, optimizing public transport network design, and facilitating the arrangement of public transport timetables. As cellular data allow us to analyze travel behaviors of the whole population, we study the problem of identify metro trip from cellular trajectory data in this work. Existing identification approaches designed for GPS trajectories often lack applicability to cellular data due to their lower spatial-temporal granularity compared to GPS data. Consequently, most approaches for cellular data focus on identifying coarse-grained transportation modes. To address the challenges of location noise and irregular data sampling for cellular data, we propose an efficient and accurate approach for fine-grained metro trip identification. Our approach employs a series of data processing operation to enhance the quality of cellular data, mitigating the effects of noise and irregular sampling. Moreover, considering the spatial-temporal information in the trajectory and metro network, we propose an effective spatial-temporal Hidden Markov Model to jointly split trajectories and identify metro trips from cellular data. Extensive experiments on two real datasets show that our proposed approach is significantly accurate and efficient for metro trip identification.