Leveraging Artificial Intelligence and Data Science for Enhancing Occupational Safety: A Multidisciplinary Approach to Risk Prediction and Hazard Mitigation in the Workplace

Abstract

The issue of occupational risks in the mining industry in Kogi State is a serious concern due to the lack of proactive safety systems. As a practical application of Artificial Intelligence (AI) and machine learning, this study aimed to forecast hazards in the workplace and use the information to develop data-driven interventions. Researchers applied the supervised models Random Forest, Support Vector Machine (SVM), Artificial Neural Network (ANN) and Decision Tree to analyze the data of 1,200 mining workers, real-time environmental sensors (PM2.5, CO, noise, temperature, vibration), and five years (2019-2024) of institutional incident records involving a total of 1,780 reported accidents. The predictive accuracy was found to be the highest in Random Forest (91.3%), ANN (88.6%), SVM (86.1%), and Decision Tree (82.5%). Random Forest model showed the best results with a precision of 0.92, a recall of 0.87, an F1-score of 0.89, and an AUC-ROC of 0.94. The best risk factor predictors were PM2.5 exposure (importance score: 0.118), PPE usage (0.105), noise levels (0.098), job role type (0.093), and CO concentration (0.089). The environmental measurements indicated that PM2.5 concentrations went up to 109µg/m3, which is significantly higher than the WHO guideline value of 25µg/m3, and noise up to 89.2dB, surpassing the OSHA limit of 85dB. Afternoon shifts recorded a significant risk of incidents, with underground drillers recording a risk rate of 54.8% against 31.6% in the morning. This project introduces Nigeria’s first verified AI model for mining, urging environmental surveillance, AI redesign, and predictive analytics in sub-Saharan occupational safety policies.