A systems-theoretic approach using association rule mining and predictive Bayesian trend analysis to identify patterns in maritime accident causes

Abstract

Accident investigations are commonly conducted to improve safety in ship design and operations. Given the lack of comprehensive approaches to understand causal factors of maritime accidents considering systems-theoretic views on accident causation, this paper presents a novel approach using information from accident investigation reports to this effect. The proposed approach combines key elements of the Causal Analysis based on Systems Theory method, Association Rule Mining and predictive Bayesian trend analysis to gain deeper understanding of patterns and trends in accident causal factors. This new approach goes beyond the state of the art by offering insights on accident causal patterns and trends at the system level, which can be used by maritime authorities and industries to enhance maritime safety by understanding co-occurring accident causes. Additionally, the approach is applied to 30 years of Canadian shipping accident reports from the Transportation Safety Board, producing new knowledge about accident causes across different commercial vessel types and accident categories. The results highlight accident causes in interactions between shipping management and vessels, and between ship crews and bridge equipment. Differences between passenger and cargo vessels, and between onboard fires and navigational accidents are observed. Discussions on results, limitations, and future research directions conclude the article.