Predicting the risky encounters without distance knowledge between the ships via machine learning algorithms

Abstract

As the maritime traffic is getting denser, the number of encounters is increasing. The aim of this study is to develop a prediction model to classify encounters as risky or non-risky when two ships encounter in a certain buffer zone. A novel methodology is proposed to integrate three-dimensional clustering in the algorithm training process. K-means clustering, and ensemble machine learning algorithms-based prediction framework is developed to overcome class imbalance. The methodology is tested in the Strait of Istanbul (SOI) and parameters are generated from a long-term AIS dataset. Framework is validated via cross validation techniques. Precision, Recall, Accuracy and ROC-AUC Score are used as measures to evaluate models. Benchmark models are generated, and the most advanced model successfully predicts each 4 out of 5 risky encounters without the knowledge of distance between two ships. Eliminating distance from decision factors provides an action period before risky encounters. Therefore, proposed framework can be a guide for autonomous vessels for safe navigation and maritime authorities to improve maritime safety.