A Bayesian network–based risk analysis approach is proposed to analyse the risk factors influencing maritime transport accidents. Comparing with previous studies in the relevant literature, it reveals new features including (1) new primary data directly derived from maritime accident records by two major databanks Marine Accident Investigation Branch and Transportation Safety Board of Canada from 2012 to 2017, (2) rational classification of the factors with respect to each of the major types of maritime accidents for effective prevention, and (3) quantification of the extent to which different combinations of the factors influence each accident type. The network modelling the interdependency among the risk factors is constructed by using a naïve Bayesian network and validated by sensitivity analysis. The results reveal that the common risk factors among different types of accidents are ship operation, voyage segment, ship type, gross tonnage, hull type, and information. Scenario analysis is conducted to predict the occurrence likelihood of different types of accidents under various situations. The findings provide transport authorities and ship owners with useful insights for maritime accident prevention.

提出了一种基于贝叶斯网络的风险分析方法，以分析影响海上交通事故的风险因素。与相关文献中的先前研究相比，它揭示了一些新功能，其中包括：（1）2012年至2017年期间，加拿大两个主要数据库海上事故调查处和加拿大运输安全委员会直接从海事事故记录中提取了新的原始数据；（2）合理分类有效预防的与每种主要海事事故类型有关的因素，以及（3）量化不同因素组合对每种事故类型的影响程度。使用朴素的贝叶斯网络构建了对风险因素之间的相互依赖性进行建模的网络，并通过敏感性分析进行了验证。结果表明，不同类型事故之间的共同危险因素是船舶运营，航次，船舶类型，总吨位，船体类型和信息。进行情景分析以预测在各种情况下不同类型事故的发生可能性。这些发现为运输当局和船东提供了预防海上事故的有用见解。