Ship Arrival Time Prediction and Berth Optimization Scheduling Based on Long Short-Term Memory Network

In order to meet the urgent needs of the maritime industry for accurate ship arrival time prediction and optimized berth scheduling, this study proposes an advanced algorithm framework based on Long Short-Term Memory (LSTM) network. The framework aims to improve the efficiency of port operations by improving the accuracy of vessel arrival time forecasts to optimize berth resource allocation. First, we pre-process the historical ship trajectory data to ensure that the LSTM model receives high-quality data input. Next, a multi-layer LSTM network structure is constructed, which is good at dealing with long-term dependencies in time series data, and by introducing three gating mechanisms, namely forgetting gate, input gate and output gate, the gradient vanishing or explosion problem is effectively solved, so as to ensure the high accuracy and stability of the model under the influence of complex factors. In order to make the model more focused on key time nodes and enhance the ability to learn important features, we integrated the Attention Mechanism into the LSTM, which allows the model to dynamically adjust the importance weights at different time points. Finally, the output of the LSTM-based prediction model is integrated into the Mixed Integer Linear Programming (MILP) model, and the optimal berth allocation strategy is determined with the objective function of minimizing the total waiting time and maximizing the berth utilization. This framework is not only superior in theoretical innovation, but also has remarkable practical application effects, providing strong technical support for the construction of smart ports and the efficient operation of the maritime industry.