Details

Machine Learning-Based Detection of Anomalous Vessel Behavior in AIS Data

Year: 2026

Term: Winter

Student Name: Jiejia Zhou

Supervisor: Michel Barbeau

Abstract: The Automatic Identification System (AIS) is a vessel tracking system that broadcasts information such as ship location, speed, and direction for navigation and maritime monitoring. As large amounts of AIS data have become available, it has become increasingly important to explore methods that can help detect unusual vessel behaviour more efficiently. This project studies how machine learning can be used to support anomaly detection in AIS vessel data. In this project, two different methods were implemented and compared. Method A uses Isolation Forest (iForest) to perform anomaly scoring on vessel movement features, and then applies a later script-based step to organize the detected anomalies into clearer categories, while Method B uses a Long Short-Term Memory (LSTM) model to evaluate vessel records across comparable dimensions. To support both methods, the AIS data was cleaned and preprocessed before analysis. The outputs of the two methods were presented using both top-ranked samples and randomly selected samples. This report describes the dataset, preprocessing steps, the design of Method A and Method B, and the way their outputs were examined and compared. Through this process, the project provides a clearer view of how different machine learning methods can be used in maritime anomaly detection.