A machine learning-based data-driven method for risk analysis of marine accidents

Abstract

In view of the frequent occurrence of marine accidents and the complex interaction of various risk-influencing factors (RIFs), a data-driven method to risk analysis that combines association rule mining (ARM) and complex network (CN) analysis is proposed in this study. The efficient FP-Growth algorithm is applied to facilitate ARM to examine risk patterns that frequently occur in marine accidents. Subsequently, CN theory is employed to scrutinise the multifaceted role of various RIFs and their interactions in the complex marine accident system, which involves the basic characteristics of the network, the identification of key RIFs through the application of the weighted LeaderRank (WLR) algorithm, and a robustness analysis. The results of the study indicate that compared with random networks, marine accident networks exhibit a higher level of complexity, which brings challenges to safety prevention and control. Inadequate regulation, violations, and deficiencies in safety management systems are identified as key RIFs, stressing the urgency of improving supervision, strengthening law enforcement and strengthening the safety management system. This study may facilitate maritime safety management of maritime traffic and the development of risk analysis methods.