Uncovering the determinants of ship detention duration in the Black Sea Region: a hybrid Bayesian and association rule mining perspective

Abstract

This study examines the key determinants of ship detention duration under Port State Control (PSC) within the Black Sea Memorandum of Understanding (MoU) region, comparing pre-pandemic and pandemic inspection outcomes. Using 756 PSC detention cases retrieved from the Black Sea Information System, a Tree-Augmented Naïve Bayesian Network (TAN) and Association Rule Mining (ARM) were applied to identify both probabilistic dependencies and frequent co-occurrence patterns. The TAN analysis revealed that PSC Authority exerted the strongest influence on detention duration, with its impact increasing notably during the pandemic. ISM Code-related deficiencies also became more critical, while traditional technical deficiencies such as Fire Safety and Life-Saving Appliances declined in significance. ARM results further highlighted that ISM non-compliance on older Ro-Ro vessels, were closely associated with detentions longer than two days in Russian and Romanian ports. These findings underline the necessity of authority-specific, risk-based inspection strategies and improved standardisation of ISM audits to minimise operational disruptions. This study is the first duration-focused analysis for the Black Sea MoU using an integrated TAN + ARM framework.