Data-driven risk assessment of occupational accidents in Indonesian open-pit coal mining using regression and neural network modeling.

BackgroundOccupational accidents in open-pit mining remain a significant workplace safety and public health concern due to hazardous environmental conditions, high-risk tasks, and complex human-organizational interactions. Data-driven risk modeling may enhance accident prevention and occupational health management.ObjectiveTo identify key determinants of occupational accidents in an open-pit coal mining operation and develop a predictive risk assessment model to support injury prevention and safety decision-making.MethodsA retrospective quantitative study was conducted using accident records (2017-2024) from a large open-pit mining company. Variables included accident timing, location, department, job position, age, work experience, unsafe acts, unsafe conditions, personal factors, and job-related factors. Multivariate regression and artificial neural network (ANN) modeling were applied to examine associations and predict the number of affected workers. Model performance was evaluated using correlation coefficients, mean squared error (MSE), mean absolute deviation (MAD), mean absolute percentage error (MAPE), and bias.ResultsUnsafe acts, hazardous environments, low worker motivation, and inadequate supervision were dominant risk contributors. Regression analysis showed significant associations between accident variables and injury occurrence (p < 0.05). The ANN model demonstrated strong predictive performance (overall R = 0.973; testing R = 0.992) with low error (MSE = 0.015; MAPE = 13.2%), while the linear regression model explained comparatively little variance (R2 = 0.094), indicating that occupational risk in this setting is driven predominantly by nonlinear interactions among behavioral, environmental, and organizational factors. Accidents were more frequent among operators, workers aged 26-30 years, and employees with >5 years of experience.ConclusionsIntegrating statistical and machine learning approaches provides a structured framework for identifying occupational risk patterns. Predictive modeling can support targeted supervision, behavioral safety interventions, and hazard control strategies in high-risk industrial settings.