Ship Age Clusters Do Not Predict Accident Severity: A Hybrid Multivariate Clustering–Logistic Model under Spatial Interaction Effects in Global Maritime
Year: 2026
Type: article
Abstract: Objectives, this study challenges the conventional assumption that ship age directly determines accident severity by examining whether age-based classifications meaningfully explain risk patterns in global maritime operations. Theoretical Framework, Grounded in Actor-Network Theory, maritime risk is conceptualized as a sociotechnical outcome shaped by interactions between vessel characteristics, spatial conditions, and analytical classification systems, rather than a purely technical attribute. Method, A dataset of 185 global maritime accident cases in 2025 is analyzed using a hybrid framework combining multivariate clustering and logistic regression. Ship age is operationalized into clusters, while location acts as a spatial moderator. Data consistency is rechecked prior to estimation to ensure robustness. Results and Discussion, Statistical results indicate no significant difference in ship age across severity levels (F = 0.36, p = 0.6996). However, clustering reveals structured patterns, with mid-aged vessels (15–20 years) accounting for approximately 69% of severe incidents. Logistic regression shows that location significantly influences severity outcomes (p < 0.01), while age clusters alone remain non-significant. This suggests that risk formation is non-linear and emerges from interaction effects rather than isolated variables. Implications, the findings highlight the need for spatially informed inspection policies, particularly in high-risk zones such as coastal and open sea areas. Regulatory frameworks should move beyond age-based thresholds toward interaction-based risk assessment and incorporate transparency in statistical classification systems. Originality/Value, to our knowledge, this study introduces a hybrid sociotechnical approach integrating clustering and regression to reinterpret maritime risk formation. By reframing statistical outputs as inscription processes, it advances both theoretical understanding and practical policy design in maritime safety governance.
Language: English
Subfield: Ocean Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7225628875