A data-driven Bayesian Network for risk modeling and causal analysis of global maritime accidents

Multiple factors may cause maritime accidents. Investigating the impact of risk factors on maritime accidents is imperative. This paper employs a data-driven Bayesian network approach to explore the impact of risk factors on maritime safety using a large dataset of maritime accidents. The interdependencies among risk influencing factors are modeled using a Tree Augmented Network, followed by the sensitivity analysis and model validation. The results indicate that the key risk influencing factors influencing maritime accidents mainly include ship location, type, age, gross tonnage, and deadweight tonnage. This study contributes to the prevention of specific types of maritime accidents.