基于关联规则挖掘和复杂网络理论的船舶碰撞事故影响因素分析  

Analysis of factors affecting ship collisions based on association rule mining and complex network theory

作　　者：冯胤伟 刘正江[1] 蒋子怡 夏国庆 曹宇皓 王新建[1] 王焕新[1] FENG Yinwei;LIU Zhengjiang;JIANG Ziyi;XIA Guoqing;CAO Yuhao;WANG Xinjian;WANG Huanxin(Navigation College,Dalian Maritime University,Dalian 116026,China;LOOM Research Institute,Liverpool John Moores University,London L33AF,UK)

机构地区：[1]大连海事大学航海学院,辽宁大连116026 [2]利物浦约翰摩尔斯大学LOOM研究所,英国利物浦L33AF

出　　处：《大连海事大学学报》2023年第3期31-44,共14页Journal of Dalian Maritime University

基　　金：国家自然科学基金青年科学基金项目(52101399);中央高校基本科研业务费专项资金资助项目(3132023138);大连海事大学博联科研基金项目(3132023617)。

摘　　要：为更加科学地分析船舶碰撞事故影响因素的交互关系,揭示船舶碰撞事故的演变机理,以全球船舶碰撞事故报告为依据,建立包含人为因素、船舶因素、管理因素、环境因素及事故时间等五类影响因素的船舶碰撞事故数据库;应用Apriori关联规则挖掘算法识别船舶碰撞事故影响因素间的频繁模式、关联、共现或因果关系,运用复杂网络理论将关联规则挖掘的结果可视化;应用拓扑特征分析方法,基于互信息理论的重要节点排序算法和基于边介数中心性的边排序算法实现关键影响因素和边的识别,对事故影响因素交互网络进行鲁棒性分析。结果表明,大部分船舶碰撞事故的影响因素较活跃且影响因素交互网络联系紧密,船舶吨位、船龄、航行水域等影响因素在交互信息传递时较为重要。In order to conduct a scientific analysis of interaction of accident influential factors,and reveal the evolutionary mechanisms of ship collision.Firstly,based on global reports of ship collision,a database covers five categories of influential factors,including human,ship,management,environment,and accident time is created.Secondly,the Apriori association rule mining algorithm is used to determine frequent patterns,associations,co-occurrences,and causal relationships among these influential factors.Visual representations of these results are obtained using complex network theory.Finally,the topological analysis methods,an important node sorting algorithm based on mutual information theory and an edge sorting algorithm based on the centrality of edge mediations are used to identify critical influential factors and edges within the network,and evaluate their robustness.The results show that the majority of the influential factors remain active and that the networks formed among these influential factors are closely linked.Influential factors such as ship tonnage,age of ship,and sailing waters are determined to be highly important in terms of information transfer and interaction.

关 键 词：水路运输 影响因素 关联规则 碰撞事故 复杂网络 机器学习 

分 类 号：U698[交通运输工程—港口、海岸及近海工程]