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

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.