Understanding unexplained crew disappearances at sea: A data-driven Bayesian risk assessment

Unexplained disappearance accidents represent a critical yet underexplored safety issue in the marine industry. Despite severe human, legal, and economic consequences, such events remain insufficiently reported and poorly understood due to inconsistent documentation and limited empirical research. To address this gap, this study develops a data-driven Bayesian Network model to examine risk-influencing factors associated with disappearance and overboard accidents, mapping their spatial distribution globally. To the authors’ knowledge, this represents the first systematic risk assessment of disappearance accidents. The model demonstrated high predictive accuracy (87.18%) in classifying accidents. Analysis reveals disappearances occur predominantly within the first two months after embarkation, most frequently on cruise vessels, with elevated risks in specific sea areas. Based on these findings, the study proposes evidence-based policy recommendations, including internationally harmonized reporting protocols, enhanced detection technologies, and structured psychological screening programs, contributing to improved maritime safety management and risk mitigation strategies. • First quantitative risk analysis of disappearance incidents at sea. • Half of the disappearance cases occur within two weeks after joining the vessel. • Highest risks concentrate in cruise, tanker, and bulk carrier segments. • Bayesian Network model achieved 87.18% accuracy in accident classification. • Findings support improved safety management and prevention strategies.