<![CDATA[Occupational health risks from chronic exposure to noise and airborne particulate matter remain a major concern in metal casting workplaces, especially in small-scale foundries with limited controls. The parameters measured in this study include noise exposure (Leq, dBA), particulate matter concentrations (PM₂.₅ and PM₁₀), and workers’ health symptoms. Field measurements at CV. Karya Yudita Baroqah showed exceedances of regulatory limits: noise levels in Molding and Finishing reached 89–93 dBA, and PM₂.₅ and PM₁₀ concentrations reached 72–80 µg/m³ and 155–174 µg/m³, surpassing recommended thresholds. These conditions indicate that workers are consistently exposed to hazardous environments that may lead to cumulative health impairments. This study aims to predict occupational health risk using a two-stage Random Forest model integrating environmental exposure data and workers’ symptoms. Stage-1 classified environmental risk levels with 99% accuracy, while Stage-2 predicted symptom-based health risk categories with 71% accuracy. PM₁₀ and PM₂.₅ were the strongest predictors, followed by noise intensity. The model demonstrates reliable performance and captures individual variability that traditional threshold-based assessments often overlook. The findings highlight that a combined machine-learning and HRA approach provides a practical, data-driven tool for early detection of high-risk workers and supports targeted interventions in metal casting workplaces.]]>
