Accurate traffic prediction is critical for the success of Intelligent Transportation Systems (ITS). However, existing models mostly train a model to predict a single task, which requires repeated training for multiple tasks and cannot use the correlation among tasks to improve performance. While multi-task learning methods (MTL) have been studied in other areas, applying them to traffic prediction is challenging due to the complex spatial-temporal dependencies in traffic data. A few transportation studies that have adopted MTL methods have only applied them to facilitate the prediction of a certain single task. To address this limitation, we propose a Spatial-Temporal Graph Multi-gate Mixture-of-Expert model (STGMMOE) via the MTL method to model multiple traffic prediction tasks simultaneously. By utilizing the correlation among tasks, STGMMOE outperforms single-task learning methods (STL) in terms of prediction performance. We evaluate STGMMOE on two real-world datasets for three prediction tasks (flow, speed, and congestion). Experimental results demonstrate that STGMMOE is an effective solution for traffic multi-task prediction, achieving higher accuracy than STL models.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Spatial-Temporal Graph Multi-Gate Mixture-of-Expert Model for Traffic Prediction


    Beteiligte:
    Guan, Tong (Autor:in) / Peng, Jiaheng (Autor:in) / Liang, Jun (Autor:in)


    Erscheinungsdatum :

    24.09.2023


    Format / Umfang :

    1931857 byte





    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch






    Dual Dynamic Spatial-Temporal Graph Convolution Network for Traffic Prediction

    Sun, Yanfeng / Jiang, Xiangheng / Hu, Yongli et al. | IEEE | 2022


    Traffic Flow Prediction Using Novel Spatial-Temporal Multi-Head Attention Graph Convolution Networks

    Cheng, Yuan / Peng, Cheng / Wang, Ze et al. | Transportation Research Record | 2024