VeTraSPM: Novel vehicle trajectory data sequential pattern mining algorithm for link criticality analysis

Abstract

This paper presents VeTraSPM (Vehicle Trajectory Data Sequential Pattern Mining), a novel algorithm designed to address the limitations of existing sequential pattern mining methods when applied to vehicle trajectory data. Current algorithms fail to capture essential characteristics such as directional flow on one-way roads (e.g., “AB” is valid but not “BA”), connectivity constraints at junctions, and the repetition of links within sequences. VeTraSPM overcomes these gaps by accurately extracting frequent patterns and confident rules while leveraging vertical projection for efficient memory and space management, enabling it to handle large datasets. Furthermore, the algorithm incorporates partitioning and parallelization techniques, further enhancing its scalability for real-world traffic environments. Three new metrics—FqMS, CMS, and SIS—are introduced to assess link criticality based on the consistent occurrence of links across movement patterns at various levels. The efficiency of VeTraSPM is demonstrated through a comparative analysis with baseline algorithms, showcasing its superior performance. The visualization of the proposed metrics offers valuable insights into link importance, supporting proactive traffic management strategies. A case study using real-world datasets from Luxembourg and Monaco validates its scalability and practical value in enhancing the resilience of urban traffic networks.