为探究中国沿海各类水上交通事故致因路径，提出一种基于人为因素分析与分类系统（HFACS）模型、贝叶斯网络（BN）模型和通径模型的水上交通事故致因路径分析方法。首先，基于HFACS模型，从中国沿海水上交通事故调查报告中识别出5个层级包含38种水上交通事故致因因素。其次，使用K2算法进行BN结构学习，并结合卡方检验结果和先验知识，确定BN结构。接着，通过最大期望算法进行BN参数学习。然后，运用敏感性分析方法提取各类水上交通事故致因路径。最后，利用通径模型计算各类水上交通事故致因路径系数，并检验其统计显著性。研究结果表明：诱发中国沿海碰撞、自沉、触碰、搁浅、火灾/爆炸、风灾和触礁等事故的显著致因路径分别有7条、6条、3条、7条、4条、4条和2条。从致因路径的效应水平上看，触碰事故和触礁事故的最大致因路径均为“海图资料配备不当→航行计划不当→触碰/触礁事故”。

 To explore the causal path of various water traffic accidents in China Coast, a causal path analysis method for water traffic accidents based on Human factors analysis and classification system (HFACS) model, Bayesian network (BN) model and path model was proposed. Firstly, based on the HFACS model, 5 levels including 38 causal factors for water traffic accidents were identified from the investigation report of water traffic accidents in China Coast. Secondly, the K2 algorithm was used for BN structure learning, and combined with chi-square test results and prior knowledge to determine the BN structure. Next, the maximum expectation algorithm was utilized for BN parameter learning. Then, the sensitivity analysis method was used to extract the causal path of various water traffic accidents. Finally, the path model was applied to calculate the causal path coefficients of various water traffic accidents and test their statistical significance. The results indicate that there are respectively 7, 6, 3, 7, 4, 4, and 2 significant causative pathways for accidents such as collision, sinking, contact, grounding, fire/explosion, wind strike, and stranding, in China Coast. From the perspective of effect level of causal path, the largest causal path of both contact accidents and stranding accidents is “improper allocation of chart data → improper navigation plan → contact/stranding accident”.