This paper proposes a spatial-temporal graph convolutional network (GCN) to sequence model (STG2Seq) to predict multi-step train arrival delays. First, the train delay data is preprocessed and encoded as the input of GCN layers. Second, gated graph convolutional module (GGCM) is composed of several GCN layers, and then GGCM is stacked to build long-term encoder and short-term encoder respectively. Among them, long-term encoder is designed to encode historical train delay data, and short-term encoder is introduced to derive the next-step prediction for generating multi-step prediction. Third, an attention-based output module is constructed to splice the output of the long-term and short-term encoder, and then important information is extracted based on channel-wise attention. Finally, the performance of proposed method for train delay prediction is evaluated based on train operation records from Wuhan-Guangzhou high-speed railway. The numerical results show that proposed method outperforms other baseline models.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Train arrival delay prediction based on spatial-temporal graph convolutional network to sequence model


    Beteiligte:
    Li, Jianmin (Autor:in) / Xu, Xinyue (Autor:in) / Shi, Rui (Autor:in) / Ding, Xin (Autor:in)


    Erscheinungsdatum :

    19.09.2021


    Format / Umfang :

    571361 byte




    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch




    Train Time Delay Prediction for High-Speed Train Dispatching Based on Spatio-Temporal Graph Convolutional Network

    Zhang, Dalin / Peng, Yunjuan / Zhang, Yumei et al. | IEEE | 2022

    Freier Zugriff

    Railway Delay Prediction with Spatial-Temporal Graph Convolutional Networks

    Heglund, Jacob S.W. / Taleongpong, Panukorn / Hu, Simon et al. | IEEE | 2020