In this study, an Association Rule Bayesian Networks (ARBN) method is established to investigate the impacts of external factors (i.e. environmental factors and ship factors) on human errors. The Bayesian networks have been structured by mining association rules from historical collision accident records. In addition, unlike other related studies that have introduced subjective data (i.e. expert knowledge), this study tends to build Bayesian networks with only objective data (i.e. accident data). In order to solve the problem of small sample size during conditional probability table estimating, this study constructs an environmental-human Bayesian network and a ship-human Bayesian network separately. The results reveal that visibility, location, and time of the day are the main environmental factors affecting the occurrence probability of human errors. For more complex scenarios, the probability of occurring judgment/operation errors in ship collision is 83% under the evidence of spring, daytime, and sea area. Among the ship factors, gross tonnage, ship types, and over safe speed show significant effects on occurring human errors. Large ships are more likely to occur judgment/operation errors under the high-speed situation. From the consequence perspective, negligence errors are highly associated with severe collisions.

在这项研究中，建立关联规则贝叶斯网络（ARBN）方法来研究外部因素（即环境因素和船舶因素）对人为错误的影响。贝叶斯网络是通过从历史碰撞事故记录中挖掘关联规则构建的。此外，与其他引入主观数据（即专家知识）的相关研究不同，本研究倾向于仅使用客观数据（即事故数据）构建贝叶斯网络。为了解决条件概率表估计中样本量小的问题，本研究分别构建了环境-人类贝叶斯网络和船舶-人类贝叶斯网络。结果表明，能见度、位置、和一天中的时间是影响人为错误发生概率的主要环境因素。对于更复杂的场景，在春季、白天和海域的证据下，船舶碰撞发生判断/操作错误的概率为83%。在船舶因素中，总吨位、船舶类型和超安全航速对发生的人为错误有显着影响。大型船舶在高速情况下更容易出现判断/操作错误。从后果的角度来看，疏忽错误与严重碰撞高度相关。超安全速度对发生的人为错误有显着影响。大型船舶在高速情况下更容易出现判断/操作错误。从后果的角度来看，疏忽错误与严重碰撞高度相关。超安全速度对发生的人为错误有显着影响。大型船舶在高速情况下更容易出现判断/操作错误。从后果的角度来看，疏忽错误与严重碰撞高度相关。