基于贝叶斯网络的海上交通安全

Bayesian Network Modeling in Maritime Safety

利用Logistic回归和贝叶斯网络模型,研究发达国家和发展中国家船舶尺度、船龄、船旗、船型及港口监控对船舶事故的影响,挖掘当前海上交通安全管理中存在的问题。通过Logistic回归计算各因素对船舶事故、检查次数和滞留情况影响的条件概率,并构建船舶事故的贝叶斯网络模型。在探讨各因素在不同状态下对各类事故的影响程度的同时,发现发生事故的船舶多为被港口国重点监控的船舶。这在一定程度上说明当前各谅解备忘录(Memorandum of Understanding,Mo U)制定的船舶检查制度在筛选危险船舶方面具有一定的有效性,但因缺乏一定的惩罚力度,未能从根本上降低船舶事故发生率。因此,建议加强各港口国监控数据的共享,建立一套完善的港口监控及惩罚机制,以有效保证航运安全,敦促船舶所有人提高船舶质量、改善全球海上运输安全。

This study tries to investigate how ship accidents are related to the status of the ship size,the ship age,the ship flag,the ship type and the PSC inspection both in developed and developing countries using Logistic regression and Bayesian network model. This could help discover the problems existing in current maritime safety management. By combing different data from various sources,the conditional probabilities of accidents and inspection and detention under different conditions are calculated using Logistic regression and builds the Bayesian network. Through empirical analysis,the impacts of various factors on different accidents are investigated. The results show that accident ships are usually those ships chosen to inspect by port States. This suggests the effectiveness of current selecting mechanisms of ship inspection defined by the Mo U( Memorandum of Understanding). However,it fails to reduce ship accidents thoroughly because of the lack of penalties. It is significant to enhance data sharing among port States and establish a systematic inspection and penalty system which will urge shipowners to improve ship quality and improve the safety of global maritime transportation effectively.