Uncovering systemic risk drivers in SIRE and CDI inspections: A hybrid TAN–ARM approach

AbstractSIRE and CDI inspection regimes are vital for managing risks in tanker and chemical carrier operations, yet empirical studies on inspection severity within these private vetting systems remain limited. This study proposes an integrated analytical framework combining Tree Augmented Naïve Bayes (TAN) and Association Rule Mining (ARM) to investigate factors influencing deficiency accumulation. A dataset of 615 standardized inspection records (2020–2024) is analyzed using vessel attributes, inspection characteristics, and deficiency categories, with the class variable defined as the number of deficiencies (<4 and ≥ 4). The TAN model quantifies conditional dependencies, while ARM identifies high-risk deficiency combinations and latent patterns. Results indicate that deficiency-type variables dominate outcomes, exceeding the explanatory power of static vessel attributes. Specifically, deficiencies in Cargo & Ballast Systems, General Appearance, Safety Management, and Pollution Prevention emerge as the primary drivers of high-deficiency inspections. Strong interaction effects among technical, operational, and managerial flaws highlight the systemic nature of inspection risk. Sensitivity analyses demonstrate that robust safety management and environmental compliance significantly mitigate inspection severity. These findings support risk-based inspection preparedness and system-oriented safety management strategies under SIRE and CDI regimes, providing data-driven insights for maritime stakeholders.