Research on Offshore Ship Trajectory Prediction and Accuracy Optimization Based on Dual-Layer LSTM-Transformer Model

The maritime risks faced by the shipping industry, as an essential pillar of global trade, are increasing with the number of ships. Ship trajectory prediction is crucial for navigation safety, but the current prediction accuracy still needs to be improved. In this paper, we propose a dual-layer LSTM-Transformer hybrid model that combines the long-time memory of the Long Short-Term Memory (LSTM) model and the Multi-Head Attention mechanism of the Transformer model. The model demonstrates lower error and higher ship trajectory prediction accuracy than traditional models.