Prediction of Coastal Port Throughput Based on Deep Learning

：This paper proposes a Long Short-Term Memory (LSTM) model integrated with an attention mechanism (Att-LSTM) for predicting bulk cargo throughput in ports. The model incorporates multi-source features, including the Baltic Dry Index (BDI), Gross Domestic Product (GDP), and vessel arrival volume, while leveraging the attention mechanism to enhance the representation of key time steps. Experiments based on real data from a specific port demonstrate that the model’s prediction accuracy significantly outperforms baseline models. Multi-step forecasting results indicate strong trend-capturing capability, confirming the model’s robustness and practicality in short- to medium-term predictions. This study provides an efficient and interpretable deep learning solution for port throughput forecasting under complex environments.