FARM: Feature-Assisted Aggregate Route Mining in Trajectory Data

An aggregate route of a set of trajectories is the representative movement direction of the set. Existing solutions address the problem of finding representative routes by finding clusters in the data with minimum intra-cluster deviation and then deriving a simplified trajectory to represent each cluster. However, existing similarity measures for trajectories are not discriminative and are sensitive to noise. This paper presents FARM, a framework for extracting aggregate routes from trajectory data. FARM first transforms the trajectories into a feature space. Next, it applies spectral clustering to find clusters in the feature space. Finally, we find a representative route for each cluster obtained. Experimental studies demonstrate the effectiveness of the proposed method.