Ship trajectory prediction based on CNN-Transformer-BiLSTM model

To improve the accuracy of ship trajectory prediction, this study employs a deep learning–based hybrid algorithm that learns motion features from Automatic Identification System (AIS) data. A hybrid CNN–Transformer–BiLSTM model is developed, integrating Convolutional Neural Networks (CNN), the Transformer architecture, and Bidirectional Long Short-Term Memory (Bi-LSTM). In this framework, CNN extracts local spatial features, the Transformer models global dependencies, and Bi-LSTM captures temporal dynamics. A representative ship in the Lianyungang port area is selected for comparative trajectory prediction experiments. Compared with BP, SVR, Bi-LSTM, and Transformer-BiLSTM, the CNN-Transforme-BiLSTM achieves significantly lower prediction errors in terms of mean squared error, root mean squared error, and mean absolute error. The experimental results demonstrate that the proposed model achieves high accuracy and strong generalization in maritime scenarios, highlighting its potential for practical traffic flow modeling.