Abstract
The integration of laser-induced breakdown spectroscopy (LIBS) with advanced machine learning algorithms, particularly extreme gradient boosting (XGBoost), offers an effective and interpretable framework for identifying transition minerals and analysing their compositions. By modelling the non-linear relationships and spectral interferences inherent in LIBS data, XGBoost significantly improves classification accuracy, robustness, and computational efficiency. The incorporation of dimensionality reduction, feature selection, and lithological context further enhances interpretability, enabling the identification of key spectral features critical for mineralogical differentiation. Model evaluations demonstrate high predictive performance for most elements – including Li, Ce, Nd, Pr, Eu, Er, Yb, Y, and Sc – evidenced by strong R2 values and low mean squared errors. In contrast, moderate accuracy for La, Sm, Gd, and Dy, and poor performance for Lu underscore the need for targeted model refinement. Overall, the results validate the potential of machine learning-based ensemble methods to advance LIBS-driven geochemical quantification, offering scalable, real-time solutions for automated, high-throughput elemental analysis in mineral exploration and sustainable resource development.
This is a visual representation of the abstract.
Get full access to this article
View all access options for this article.