Enhanced prediction of occurrence forms of heavy metals in tailings: a systematic comparison of machine learning methods and model integration

Tailings produced by mining and ore smelting activities are a major source of soil pollutants. Understanding the speciation of heavy metals (HMs) in tailings is essential for soil remediation and sustainable development. Given the complex and time-consuming nature of traditional laboratory sequential extraction measurement methods for determining the occurrence forms of HMs in tailings, a rapid and precise identification approach is urgently required. To address this issue, in this study, a general empirical prediction method for the occurrence forms of HMs was developed using machine learning (ML). The compositional information of the tailings, the properties of the HM elements, and the sequential extraction steps were used as inputs for calculating the percentages of seven occurrence forms of HMs. After model tuning and comparison, extreme gradient boosting, gradient boosting decision tree, and categorical boosting methods were found to be the three best-performing ML models, with R2 values on the testing set exceeding 0.859. The feature importance analysis for these three optimal models indicated that electronegativity was the most important factor affecting HM occurrence, with an average feature importance of 0.452. Subsequently, stacking was used as a model integration method to further improve the ML models’ ability to predict HM occurrence forms, increasing the R2 to 0.879. Overall, this study presents a robust technique for predicting HM occurrence forms in tailings and provides an important reference for the environmental assessment and recycling of tailings.