Građevinar, Vol. 78 No. 04., 2026.
Original scientific paper
https://doi.org/10.14256/JCE.4514.2026
Predicting the probability of occupational accident outcomes in the construction industry
Şükrü Bulut
Muhammed Furkan Kahraman
Serap Yörübulut
Ahmet Kürşad Türker
Adnan Aktepe
Abstract
The construction industry experiences numerous accidents of varying severity. This study analyses 434,134 occupational accident records from the construction sector in Turkey (2012–2021) using ordinal logistic regression for four-category severity modelling. Unlike previous studies, the dataset was split into 70 % training and 30 % independent test sets. The model achieved a classification accuracy of 56.0 %, demonstrating strong sensitivity in identifying rare, high-severity outcomes, such as death and permanent disability. These results provide a validated predictive equation and identify critical risk factors, thereby offering a data-driven framework for prevention strategies. This comprehensive modelling approach bridges the gap between theoretical regression and practical safety decision-making.
Keywords
construction; occupational health and safety; occupational accident; ordinal logistic regression; risk factors; accident severity; predictive modelling
Hrčak ID:
348647
URI
Publication date:
9.6.2026.
Visits: 0 *