Vessel classification using AIS data

Maritime Domain Awareness (MDA) relies heavily on Automated Identification System (AIS) data for vessel tracking. This research focuses on developing a novel vessel classification framework that uses AIS derived features. The algorithm effectively classifies ocean-going vessels into behavioural categories, providing valuable insights for MDA. demonstrate the effectiveness of the classification framework in achieving high accuracy (F1 score of 0.88–0.9) in vessel classification. The choice of class labels and data pre-filtering significantly impacts performance. The algorithm's feature importance analysis highlights the relevance of self-reported vessel dimensions, location, and behaviour. While cargo and tanker vessels exhibit some overlap, fishing vessels are accurately classified. However, recreational and passenger vessels, due to limited samples, require further refinement. Future research could explore time series methods and tailored algorithms for specific vessel classes to enhance classification accuracy. Overall, this study contributes to improving MDA by providing a robust vessel classification tool. Further investigation is needed to address the high proportion of unlabeled vessels classified as fishing vessels. • Large scale, high accuracy ocean vessel classification. • LightGBM used with AIS data. • AIS trajectory features used in machine learning. • Global AIS dataset used in for machine learning.