A taxi detour trajectory detection model based on iBAT and DTW algorithm

Taxi detour is a chronic problem in urban transport systems, which largely undermines passengers' riding experience and the city's image while unnecessarily worsening traffic congestion. Tourists unfamiliar with city roads often encounter detour problems. Therefore, it is important for regulatory authorities to develop a tool for detour behavior detection in order to discover or identify detours. This study proposes a detour trajectory detection model framework based on the trajectory data of taxis that can identify taxi driving detour fraud at the microscopic level and analyze the characteristics of detouring trajectories from the perspective of microscopic motion traits. The deviation from normal driving trajectories provides a framework for the automatic detection of detour trajectories for the off-site supervision platform of the taxis. Considering drawbacks of the isolation-Based Anomalous Trajectory (iBAT) algorithm, this paper made further improvements in trajectory anomaly detection. In this study, three methods including the iBAT, iBAT + Dynamic Time Warping (DTW), and iBAT + DTW algorithms considering the driving distance and time are compared using the relevant experimental data. The case studies verify that the proposed method outperforms the other methods. Verified by the experiments based on the trajectory data coming from Nanjing, the false positive rate of this framework is only 1.64%.