Exploring urban travel patterns using density-based clustering with multi-attributes from large-scaled vehicle trajectories

Abstract Extracting travel patterns from large-scaled vehicle trajectories is the key step to analyze urban travel characteristics, which can also provide effective strategies for urban traffic planning, construction, management and policy decision. In this study, we adopt the DBSCAN (Density-Based Spatial Clustering of Application with Noise) algorithm by fusing spatial, temporal and directional attributes extracting from vehicle trajectories Furthermore, LCS (Longest Common Sequences) is adopted to estimate spatial similarity, and two measurements are also designed to evaluate the temporal and directional similarity between trajectories. Accordingly, a statistical feature-based parameter optimization method is proposed in the clustering process to achieve reasonable clustering results. Finally, trajectory data collected from Harbin city, China are used to validate the effectiveness of clustering method. A comparison of clustering results considering different combination of attributes is conducted to further demonstrate the advantage of the proposed model.