Navigational accidents (collisions and groundings) account for approximately 85% of mari-time accidents, and consequence estimation for such accidents is essential for both emergency resource allocation when such accidents occur and for risk management in the framework of a formal safety assessment. As the traditional Bayesian network requires expert judgement to develop the graphical structure, this paper proposes a mutual information-based Bayesian network method to reduce the requirement for expert judgements. The central premise of the proposed Bayesian network method involves calculating mutual information to obtain the quantitative element among multiple influencing factors. Seven-hundred and ninety-seven historical navigational accident records from 2006 to 2013 were used to validate the methodology. It is anticipated the model will provide a practical and reasonable method for consequence estimation of navigational accidents.

中文翻译：

---

基于互信息的长江航行事故后果估计贝叶斯网络模型

航行事故（碰撞和搁浅）约占海上事故的 85%，此类事故的后果估计对于此类事故发生时的应急资源分配和正式安全评估框架内的风险管理都是必不可少的。由于传统的贝叶斯网络需要专家判断来开发图形结构，本文提出了一种基于互信息的贝叶斯网络方法来降低对专家判断的要求。所提出的贝叶斯网络方法的中心前提涉及计算互信息以获得多个影响因素之间的定量元素。使用 2006 年至 2013 年的 797 条历史航行事故记录来验证该方法。