Vessel Classification from Space-based AIS Data Using Random Forest

Abstract

AIS data provides abundant near real-time information about the vessel of interest for maritime traffic management as well as rich data source for machine learning based data mining. Vessel classification is considered as an important topic to improve maritime situation awareness (MSA). In this work, random forest is used to classify three major vessel types from their geometric features under different parameter settings, moreover, we also analyze the importance of the features extracted and the experiments demonstrate that random forest can produce good classification results under proper parameter setting and it still shows relatively good performance on the samples with several selected important features.