An enhanced sine cosine algorithm based light gradient boosting machine for occupation injury outcome prediction

Occupational injuries pose significant risks to workforce safety and productivity, yet conventional predictive methods often underperform due to suboptimal parameter tuning and limited modelling of complex interactions. This study presents a hybrid Enhanced Sine Cosine Algorithm (ESCA)–Light Gradient Boosting Machine (LightGBM) framework for accurate injury outcome prediction. ESCA leverages differential mutation and local search to enhance convergence and avoid local optima. Benchmarking on CEC2015 functions and application to a socio-economic injury dataset demonstrates that ESCA-LightGBM consistently outperforms other competing models, achieving a predictive accuracy of 99.1%. Feature importance analysis identifies health expenditure, industry employment, and life expectancy as the most influential predictors. These results not only confirm the robustness of the proposed model but also highlight its potential to inform occupational health policy, optimize resource allocation, and strengthen proactive risk management strategies.