Feature Transformer and LightGBM Ensemble for Ship Trajectory Recognition Using Real AIS Data

The Automatic Identification System (AIS) generates massive volumes of real-world ship trajectory data, providing a critical foundation for maritime ship type classification. However, existing methods often struggle to simultaneously capture long-range temporal dependencies, maintain computational efficiency, and ensure model interpretability, which makes accurate multi-class classification challenging in real-world maritime environments. To address these limitations, this study proposes a robust and efficient hybrid framework. The proposed architecture integrates a Feature Transformer module for deep temporal feature extraction with a LightGBM model for efficient ensemble classification. Specifically, the multi-head self-attention mechanism within the Feature Transformer captures long-range dependencies in preprocessed AIS sequences to generate compact trajectory fingerprints. These deep temporal representations are then concatenated with carefully designed statistical and kinematic tabular features and fed into the LightGBM classifier for final ship type identification. To validate the proposed framework, we construct a comprehensive real-world AIS dataset consisting of 2,196 trajectories collected between 2019 and 2023, encompassing diverse ship types that reflect authentic maritime scenarios. Experimental results show that the proposed method achieves 82.42% overall accuracy and 77.35% Macro-F1, significantly outperforming comparative baseline models, including LSTM (64.85% accuracy), GRU (64.85%), vanilla Transformer (61.21%), and standalone LightGBM (59.09%). Furthermore, the hybrid model offers ultra-fast inference (1.58 ms per batch) and enhanced interpretability through SHAP-based analysis, making it highly suitable for near real-time maritime traffic monitoring and decision-support applications.