Scholarlink

<p>Nickel laterite deposits are compositionally complex and require reliable grade classification to support efficient exploration, ore blending, and resource management. This study evaluates the performance of four supervised machine learning algorithms—Random Forest (RF), Gradient Boosting Classifier (GBC), Support Vector Machine (SVM), and Multilayer Perceptron (MLP)—for nickel laterite grade classification using multivariate geochemical data obtained from wavelength dispersive X-ray fluorescence (WDXRF) analysis. A total of 624 elemental measurements, including Ni, Fe, Mg, Al, Cr, Si, and other associated elements, were analyzed from laterite samples collected in eastern Sulawesi, Indonesia. The samples were categorized into low-, medium-, and high-grade classes based on industrially relevant nickel concentration thresholds. Data preprocessing included cleaning, standardization, stratified training–validation–test partitioning, and five-fold cross-validation. Model performance was evaluated using accuracy, precision, recall, F1-score, confusion matrix, ROC/AUC, and Cohen’s Kappa metrics. The results show that the ensemble-based methods outperformed the other classifiers, with RF and GBC achieving perfect classification accuracy (1.00) and excellent robustness across the validation folds. SVM and MLP also demonstrated strong predictive capability, with accuracies of 0.96 and 0.94, respectively, although moderate overlap occurred within the medium-grade class due to transitional geochemical characteristics. Feature importance analysis consistently identified Ni, Mg, and Fe as the dominant class differentiation predictors, corresponding to saprolite–limonite zonation in lateritic profiles. The findings demonstrate that integrating WDXRF geochemical analysis provides a rapid, interpretable, and scalable framework for nickel-grade classification and resource evaluation. This approach offers significant potential for automated geochemical workflows and intelligent decision-making in lateritic nickel mining.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimization of Ensemble Machine Learning Models for Ore Grade Classification of Nickel Laterite Based on Multivariate Geochemical Data from the Sulawesi Ophiolite Belt, Indonesia

  • 期刊论文

  • 发表日期：

    2026年7月28日

查看全文
  • Karlina Maulida,
  • Murat Özen

摘要

Nickel laterite deposits are compositionally complex and require reliable grade classification to support efficient exploration, ore blending, and resource management. This study evaluates the performance of four supervised machine learning algorithms—Random Forest (RF), Gradient Boosting Classifier (GBC), Support Vector Machine (SVM), and Multilayer Perceptron (MLP)—for nickel laterite grade classification using multivariate geochemical data obtained from wavelength dispersive X-ray fluorescence (WDXRF) analysis. A total of 624 elemental measurements, including Ni, Fe, Mg, Al, Cr, Si, and other associated elements, were analyzed from laterite samples collected in eastern Sulawesi, Indonesia. The samples were categorized into low-, medium-, and high-grade classes based on industrially relevant nickel concentration thresholds. Data preprocessing included cleaning, standardization, stratified training–validation–test partitioning, and five-fold cross-validation. Model performance was evaluated using accuracy, precision, recall, F1-score, confusion matrix, ROC/AUC, and Cohen’s Kappa metrics. The results show that the ensemble-based methods outperformed the other classifiers, with RF and GBC achieving perfect classification accuracy (1.00) and excellent robustness across the validation folds. SVM and MLP also demonstrated strong predictive capability, with accuracies of 0.96 and 0.94, respectively, although moderate overlap occurred within the medium-grade class due to transitional geochemical characteristics. Feature importance analysis consistently identified Ni, Mg, and Fe as the dominant class differentiation predictors, corresponding to saprolite–limonite zonation in lateritic profiles. The findings demonstrate that integrating WDXRF geochemical analysis provides a rapid, interpretable, and scalable framework for nickel-grade classification and resource evaluation. This approach offers significant potential for automated geochemical workflows and intelligent decision-making in lateritic nickel mining.