Data-Augmented Maritime Accident Severity Prediction: A Port State Control Records-based Fusion Approach

Li, T, Wang, X, Chang, C-H ORCID: 0000-0002-7351-8471, Grifoll, M, Feng, H and Yang, Z ORCID: 0000-0003-1385-493X (2026) Data-Augmented Maritime Accident Severity Prediction: A Port State Control Records-based Fusion Approach. Reliability Engineering and System Safety, 278. ISSN 0951-8320

Preview

Text Data augmented maritime accident severity prediction A port state control records based fusion approach.pdf - Published Version Available under License Creative Commons Attribution. Download (6MB) | Preview

Abstract

Data-driven maritime accident analysis based on mainstream accident databases is approaching a high performance ceiling, calling for further breakthroughs from both data quality and modelling perspectives. Port State Control (PSC) inspection records provide additional pre-accident vessel safety information, offering a valuable opportunity to improve accident severity prediction. This study therefore proposes a PSC-assisted accident severity prediction framework that provides a basis for pre-accident assessment. First, a multi-source prediction database is constructed by integrating Paris MoU PSC inspection records with accident information from the Global Integrated Shipping Information System. To address the limited matched samples and high-dimensional categorical features, a two-stage augmentation strategy is developed. Finally, tree-model-based feature selection is applied, and predictive performance is evaluated. The results show that incorporating PSC data reduces the average Mean Squared Error (MSE) by 1.15%. Cross-validation further shows that the complete framework achieves an additional 1.99% reduction in average MSE compared with the PSC–accident fusion-only setting. In particular, the best-performing LightGBM model achieves an MSE reduction of 11.22% relative to the original accident-data-only baseline. This study demonstrates that PSC inspection records are a valuable supplementary data source for maritime accident severity prediction and highlights their significant role in enhancing vessel accident prediction models.

Item Type:	Article
Uncontrolled Keywords:	3505 Human Resources and Industrial Relations; 35 Commerce, Management, Tourism and Services; 01 Mathematical Sciences; 09 Engineering; 15 Commerce, Management, Tourism and Services; Strategic, Defence & Security Studies; 35 Commerce, management, tourism and services; 40 Engineering; 49 Mathematical sciences
Subjects:	H Social Sciences > HD Industries. Land use. Labor > HD61 Risk Management T Technology > TA Engineering (General). Civil engineering (General) V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering
Divisions:	Engineering and Built Environment
Publisher:	Elsevier
Date of acceptance:	18 October 2026
Date of first compliant Open Access:	30 October 2026
Date Deposited:	30 Oct 2026 11:56
Last Modified:	30 Oct 2026 11:56
DOI or ID number:	10.1016/j.ress.2026.113691
URI:	https://researchonline.ljmu.ac.uk/id/eprint/29783

View Item