Search

Accurate forecasting of vessel traffic flow is crucial for establishing a smart inland waterway transportation system, ensuring real-time safety supervision, and enabling prompt emergency response. While existing methods mainly focus on temporal correlation, there is a need for research on utilizing spatiotemporal features for area vessel traffic forecasting. The study proposes a novel spatiotemporal vessel traffic fluxes prediction model called STA-BiLSTM. The model combines the Graph Attention (GAT) network and Long Short-Term Memory (LSTM) network within the GAT-LSTM framework to capture the spatial and temporal aspects of vessel traffic patterns. The proposed model incorporates the attention mechanism and Bi-directional Long Short-Term Memory (Bi-LSTM) model to enhance the accuracy and robustness of ship traffic flow prediction by capturing the dependencies between time steps and waterway segments. Experimental results based on three actual Yangtze River vessel traffic flow datasets demonstrate the superiority of the proposed method over traditional approaches. Furthermore, ablation experiments validate the effective extraction of spatiotemporal dependency features between river segments by the STA-BiLSTM model. Finally, the paper showcases the significance of the proposed method through practical examples. • A new graph organizational form is used to inland waterway transport network. • GAT-LSTM framework performs the LSTM and GAT methods in inland waterway transport network. • STA-BiLSTM model is presented to deal with the vessel traffic flow forecasting problem. • The performance of the STA-BiLSTM model is higher than other competitors, which demonstrates the superiority of our model. [ABSTRACT FROM AUTHOR]