Assessing the accident severity level of passenger vessels in Indonesia using Bayesian Network model

The growing demand for passenger vessels has been paralleled by increased accidents, resulting in significant economic, human, and environmental losses. Accidents on passenger ships often stem from complex factors, including technical, operational, and human elements. Therefore, a detailed analysis is essential for understanding these factors and improving safety management. While various traditional risk analysis methods exist, the Bayesian Network (BN) offers unique advantages in modelling the probabilistic relationships between risk factors and accident outcomes. This study aims to analyse the accident severity level of passenger vessels in Indonesia by employing a Tree Augmented Naïve Bayesian Network (TAN-BN) to assess 46 passenger ship accidents in Indonesia using 17 identified Risk Influencing Factors (RIFs) focused on ship internal factors. Sensitivity analysis using mutual information and True Risk Influence (TRI) methods identified “Ship Operation” and “Accident Type” as the most significant RIFs, where the ship during passage is the most severe ship operation, and the ship sinking accident is the most catastrophic accident type. Scenario analysis revealed that very serious accidents often occur in transit, with human factors, particularly violation errors, playing a critical role. This study can leverage the decision-making process for stakeholders to reduce the severity of accidents in passenger vessels.