Incorporating human and organizational failures into the formation pattern for different Arctic maritime accidents using a data-driven Bayesian network.

To gain a comprehensive understanding of formation patterns for various maritime accidents in Arctic waters, the present study integrates human and organizational failures (HOFs) with accident risk factors (AIFs) to examine the individual or collective impacts of different AIFs and HOFs on Arctic maritime accidents. Firstly, a novel database is established within the improved Human Factors Analysis and Classification System (HFACS), encompassing 14 AIFs with multiple states and 33 HOFs with binary states. Secondly, the Association rule mining (ARM) is employed to capture the causal relationships among HOFs, while Augmented Naive Bayes (ANB) is introduced to account for the interdependency of AIFs, thereby developing the data-driven BN model for analyzing and predicting Arctic maritime accidents. Finally, the criticality and sensitivity of AIFs and HOFs across different Arctic maritime accidents are identified and ranked by BN forward and backward inference. Furthermore, three scenarios are simulated to demonstrate effective management strategies for prevention and mitigation of Arctic maritime accidents. Additionally, a correlation graph depicting the relationship between HOFs and AIFs is generated for each accident scenario, facilitating crew onboard ships as well as shipping companies and maritime authorities in customizing targeted safety measures aimed at preventing Arctic maritime accidents. • A comprehensive database encompassing HOFs and AIFs associated with Arctic maritime accidents is established. • A data-driven approach integrating improved HFACS, ARM, and data-driven BN is proposed. • The impact of different AIFs and HOFs individually or collectively on Arctic maritime accidents is investigated. [ABSTRACT FROM AUTHOR]