Real-Time Detection of Anomalous Taxi Trajectories from GPS Traces

Abstract

Trajectories obtained from GPS-enabled taxis grant us an opportunity to not only extract meaningful statistics, dynamics and behaviors about certain urban road users, but also to monitor adverse and/or malicious events. In this paper we focus on the problem of detecting anomalous routes by comparing against historically “normal” routes. We propose a real-time method, iBOAT, that is able to detect anomalous trajectories “on-the-fly”, as well as identify which parts of the trajectory are responsible for its anomalousness. We evaluate our method on a large dataset of taxi GPS logs and verify that it has excellent accuracy (AUC ≥ 0.99) and overcomes many of the shortcomings of other state-of-the-art methods.