A Bayesian Network for Classifying and Predicting Ship Collision

Ship collisions are the type of accident with the highest percentage of investigations, making them the type of accident with a high variation in causes. Additionally, ship collisions pose a serious threat because they occur between two different vessels, resulting in material losses and loss of life. This condition makes ship collisions a serious problem that requires efforts to minimize prevention and adjust existing conditions. This study aims to model the causes of ship collisions in Indonesia to determine the probability of a ship experiencing a collision or a near miss. The modeling will be conducted using the Bayesian network method. The Bayesian network model is based on the factors that cause ship collisions, relying on past incidents and written reports from National Transportation Safety Committee (NTSC) investigations and judgments from the Maritime Court. The purpose of this study is to identify the factors that cause ship collisions, determine the probability of a ship experiencing a collision, and identify the factors that contribute the most to the probability of ship collisions in Indonesia through sensitivity analysis. The results obtained from the model, with a 70% weight for training data, show that the probability of a ship experiencing a collision during a dangerous condition is 63%, with an accuracy and sensitivity of 93.75% and 100% respectively. According to the model, the factors with the greatest influence are “crew competence,” “decision making,” “maneuverability,” and “ship communication.”