Research on Optimization and Application of Ship Trajectory Prediction Method Based on BiLSTM-Attention

The task of ship trajectory prediction is to accurately forecast the future position of a ship. To address the limitations of traditional models in handling time dependencies and long-term prediction accuracy, this paper proposes a ship trajectory prediction model based on BiLSTM-Attention. The model combines Bidirectional Long Short-Term Memory (BiLSTM) to extract temporal features and introduces an Attention Mechanism to weight key time steps, optimizing the prediction results. Additionally, a Multi-head Self-Attention mechanism is employed to address long-range time step dependencies. Experimental results demonstrate that the proposed BiLSTM-Attention model significantly outperforms both LSTM and BiLSTM in multiple performance metrics, validating its effectiveness in ship trajectory prediction tasks. This model provides strong support for navigation safety and the application of intelligent shipping technologies.