Machine learning-based prediction of heavy metal leaching efficiency from zinc metallurgical waste

The increasing generation of hazardous industrial residues containing critical metals requires efficient resource-recovery strategies that simultaneously mitigate environmental burdens and promote resource circularity. Herein, an integrated hydrometallurgical-machine learning framework was developed for the recovery and predictive modeling of gallium and germanium from zinc plant residues. Key process variables, including acid concentration, leaching temperature, and reaction time, were employed as model inputs to predict metal leaching efficiency. Four supervised learning algorithms, namely support vector machine, random forest, extreme gradient boosting, and artificial neural network, were developed and systematically evaluated using a 60:40 training-test data partition, with model robustness further assessed through 10-fold cross-validation. Among the evaluated models, XGBoost exhibited the highest predictive performance for Ga leaching efficiency, with coefficients of determination (R 2 ) of 0.94 for the training set and 0.81 for the independent test set, demonstrating satisfactory generalization to unseen experimental conditions. Integrating data-driven prediction with hydrometallurgical processing provides a quantitative framework for reducing experimental requirements, identifying influential process variables, and improving process predictability. The proposed approach offers a scalable pathway for valorizing metal-bearing industrial residues while supporting sustainable waste management and the circular utilization of critical resources.