Machine learning-based classification of maritime accident severity: a comparative approach
DIGITAL TRANSPORTATION AND SAFETY, cilt.5, sa.2, ss.170-181, 2026 (ESCI, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 5 Sayı: 2
- Basım Tarihi: 2026
- Doi Numarası: 10.48130/dts-0026-0014
- Dergi Adı: DIGITAL TRANSPORTATION AND SAFETY
- Derginin Tarandığı İndeksler: Emerging Sources Citation Index (ESCI), Scopus
- Sayfa Sayıları: ss.170-181
- İstanbul Üniversitesi Adresli: Evet
Özet
The continuous growth of global trade has significantly increased maritime traffic density and navigational complexity, escalating the risk of maritime accidents and their catastrophic consequences. While traditional research has focused on accident frequency, accurately predicting accident severity is vital for enhancing safety protocols and optimizing emergency resource allocation. This study aims to predict the severity of maritime accidents by using classification-based machine learning algorithms. This study proposes a comprehensive comparative framework for classifying maritime accident severity using an open-access dataset containing 223 accidents recorded between 2015 and 2020, incorporating 18 risk-influential variables related to vessel characteristics and environmental conditions. The research evaluates three primary machine learning algorithm families: Ensemble Trees, Support Vector Machines, and Neural Networks. Model performance is rigorously compared using metrics such as accuracy, total cost, error rate, precision, recall, and F1-score to identify the most robust predictor. This comparative approach addresses the limitations of traditional statistical methods in handling non-linear, high-dimensional maritime data. As a result, the most successful algorithm was the Bagged Tree algorithm with an accuracy rate of 86.4%. The attributes that are effective in the decision mechanism of the model, especially the effect of variables such as the number of casualties, flag state, total length, sea area of occurrence, and visibility distance, are emphasized. The findings provide stakeholders with a robust scientific basis for proactive risk assessment and targeted accident prevention strategies, ultimately contributing to safer global maritime operations.