A hybrid model for marine accident analysis based on Bayesian Network (BN) and Association Rule Mining (ARM)

In order to ensure sustainable maritime safety, studies based on unreported maritime accidents in maritime transport are necessary. Such studies allow the causes of accidents that have not come to light, to be identified and addressed. In this study, the data of unreported occupational accidents on Turkish fishing vessels with a full length of 12 m and above was analysed using both Bayesian network (BN) and Association Rule Mining (ARM) methods. A network structure that summarizes the occurrence of occupational accidents on fishing vessels with the BN method was put forward. The network structure makes it possible to analyse the latent factors, active failures and operational conditions that cause the accident qualitatively and quantitatively. The Predictive Apriori algorithm was used to establish rules for the occurrence of occupational accidents on fishing vessels, taking variables such as day condition, length, sea condition, and ship type into account. These rules provide an understanding of how occupational accidents occur on fishing vessels. In other words, these rules define the minimum requirements for the occurrence of accidents on fishing boats. The developed hybrid model can be used for analysing unreported occupational accidents on fishing vessels.

Automated summary

In order to ensure maritime safety, it is necessary to study unreported maritime accidents. This study uses Bayesian network and Association Rule Mining methods to analyze data of unreported occupational accidents on Turkish fishing vessels. A network structure and accident occurrence rules are proposed to help analyze the latent factors and requirements for occupational accidents on fishing vessels.