A novel object-oriented Bayesian network on risk assessment of sea lanes of communication

Abstract

The maritime transportation system heavily depends on sea lanes of communication. Unfortunately, the surge in shipping activities has also resulted in an increase in maritime incidents, encompassing both conventional accidents and piracy-related events. This paper presents a novel object-oriented Bayesian network framework for assessing the risk of traditional maritime accidents and piracy-related incidents. Furthermore, an enhanced machine-learning-based Tree augmented naïve learning approach is proposed to address the correlation relationships among the influential factors and the distinctions between mathematical and practical meanings. The paper also introduces mutual information theory and an improved scenario construction approach to identify the key influential factors on maritime risks and unveil the non-linear relationships among them. The findings indicate that both the degraded natural environment and social environment contribute to the risk levels of maritime accidents. Furthermore, the type of ship is recognized as the most significant factor contributing to unsafe conditions. The research results will offer valuable insights to the maritime industry, specifically aiding ship owners, ship crew, and shipping companies in understanding the key influential factors in multiple types of maritime accidents and assessing potential risks.