ABSTRACT
Maritime accident research conducted at conventions often suffers from insufficient quantitative descriptions of the correlations between factors. This paper aims to conduct new research by incorporating maritime accident data from eight coastal areas into a data-driven Bayesian network model (BN) to analyze the relationship between the severity of marine accidents and relevant Accident Influential Factors (AIFs). To achieve this aim, the factors contributing to maritime accidents are identified, and each of them is subjected to a chi-squared test against the severity of the accident. The results of the chi-square test are then compiled into a comparison matrix, which is imported into the Bayesian network as a priori probability. Next, a Bayesian search algorithm is used to establish a data-driven BN model, and the AIFs of the database for eight different coastal areas are analyzed through data training and machine learning based on the prior probability. This study found that “Type of accident,” “Weather,” and “Sea condition” have the highest correlation with accident severity. Additionally, factors such as visibility, ship size, and age significantly impact serious accidents. The research findings are beneficial for the analysis of marine accidents and provide insights to stakeholders such as safety authorities and ship owners.
