Anomalous Trajectory Detection from Taxi GPS Traces Using Combination of iBAT and DTW

The competition between taxi companies triggers the occurrence of taxi driving frauds. The traces of driving frauds often significantly deviate from normal ones. It is possible to automatically detect the anomalous trajectory by mining historical GPS traces. Uniform grids are used to represent the trajectory in isolation-Based Anomalous Trajectory (iBAT) method and used to address the issue of GPS traces. However, the uniform grid has its own challenges where the trajectories having similar patterns can generate different sequence of grid cells. This situation allows a normal trajectory to be detected as an anomaly. In this study, we proposed the combination of iBAT and similarity measurement like DTW. By combining iBAT and DTW return the lowest false alarm rate (FAR), which is 0,027. Based on the result, the proposed method is proven to be able to minimize normal trajectory is detected as an anomaly.