Machine learning-enhanced grey wolf optimization of microwave-assisted arsenic extraction in rice: a comprehensive analytical framework

This study presents an innovative approach for optimizing microwave-assisted arsenic extraction in rice samples by integrating multiple machine learning (ML) models with grey wolf optimization (GWO). Four critical extraction parameters were systematically evaluated: microwave power (300–600 W), temperature (50–80 °C), extraction time (30–60 min), and nitric acid concentration (0.1–0.5 M). A comprehensive hybrid framework was developed, incorporating six distinct machine learning (ML) models: The following boosting methods are employed: XG Boosting, LS Boosting, KRR_RBF (Kernel Ridge Regression with RBF kernel), SVR_Poly (Support Vector Regression with Polynomial kernel), a weighted hybrid, and a stacked ensemble. The stacked ensemble model achieved excellent performance (R²=0.986, CCC = 0.993, EF = 0.986) and outperformed individual models. GWO optimization identified three optimal operating points, with the highest yield (1.21 µg/g) achieved at 599.52 W and 50.03 °C, exhibiting only 4.20% prediction error. Alternative configurations at medium power (525 W) and low power (474.99 W) demonstrated process flexibility while maintaining prediction errors below 8%. The proposed method exhibited a low LOD (0.03 µg/Kg). This study introduces a new method for optimizing analytical chemistry processes using machine learning and optimization algorithms. This approach enhances prediction accuracy and provides valuable operational guidelines for arsenic extraction, with potential for broader applications in analytical chemistry requiring complex parameter optimization.