An Estimation of Ship Collision Risk Based on Relevance Vector Machine

According to the statistics of maritime collision accidents over last five years (2016–2020), 95% total are caused by human factors. Machine learning algorithms an emerging approach in judging risk among vessels and supporting reliable decision-making prior any behaviors for avoidance. As result, it can be a good method reduce errors navigators’ carelessness. This article aims propose enhanced machine estimate ship support more risk. In order risk, conventional vector (SVM) was applied. Regardless advantage SVM resolve uncertainty problem using collected ships’ parameters, has inherent weak points. this study, relevance (RVM), which present probabilistic results based on Bayesian theory, applied The proposed compared with applying SVM. It showed that estimation model RVM is accurate efficient than We expect reasonable navigator through estimation, thus allowing early evasive actions.