A Statistical Method for Detecting Move, Stop, and Noise: A Case Study with Bus Trajectories
The proliferation of devices with positioning capability has allowed new possibilities for studies and applications in the context of urban mobility. However, the process of analyzing raw trajectories poses several challenges. In this work, we investigate one of the main tasks in this process of trajectory analysis: detecting stops from GPS trajectories. Stops can reveal interesting behavior aspects of a moving object such as its daily routine, bottlenecks in traffic jams, or visiting times of touristic places. Although there are some efforts in this direction, most current methods ignore the presence of noise segments, which typically occur many times in trajectories. In this sense, we present a method that exploits gaps in time and space to identify episodes of movement, stop, and periods where some classification is inconclusive, which we define as noise. In addition, our method does not rely on contextual information as opposed to some current solutions, which make our proposal also suitable for trajectories recorded in free space. We compare our method to the state of the art highlighting its advantages in terms of manipulating noise, supporting spatial filtering and being independent of external resources. Moreover, we conduct an experimental evaluation using a large-scale bus dataset to show the effectiveness of our method in a real application scenario.