Scholarlink

<p>Galena, a primary lead ore, is found in significant deposits globally, including sediment-hosted lead–zinc, volcanogenic massive sulfide, porphyry copper, skarn, and epithermal gold deposits, often associated with valuable metals such as Ag, Sb, Tl, and Sn. The classification of galena based on trace elements is crucial for understanding metallogenic origins and improving mineral exploration strategies. This study utilized machine learning algorithms, including random forest (RF), gradient boosting (GB), multilayer perceptron (MLP), and support vector machine (SVM), to classify galena samples from 37 Pb–Zn deposits worldwide. To address dataset imbalance, resampling techniques including the synthetic minority oversampling technique (SMOTE) and random undersampling with clustering (RUC) were employed to enhance model performance and reduce bias. K-fold cross-validation (k = 10) was implemented to ensure model robustness. RF and GB achieved the highest accuracy, 97.62% and 98.19%, respectively, on oversampled data, while MLP and SVM demonstrated lower sensitivity to resampling, with SVM performing least effectively on undersampled data (86.65% accuracy). Key trace elements, such as Sn, Tl, and Ag, were identified as significant discriminators across deposit types, offering insights into ore-forming processes. Additionally, metamorphic recrystallization in galena samples, as indicated by trace element redistribution under varying geochemical conditions, provides further metallogenic insights. The use of t-distributed stochastic neighbor embedding (t-SNE) revealed distinct clustering patterns among galena samples, validating trace element data for metallogenic classification. These findings have significant implications for mineral exploration, particularly in data-limited regions, as the integration of machine learning with trace element geochemistry provides an efficient tool for targeting Pb–Zn deposits and identifying metamorphic overprinting zones, thereby optimizing exploration efforts.</p> Graphical Abstract <p></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

Big Data Mining on Galena Geochemistry Using Machine Learning Algorithms: Implications for Metallogenic Discrimination

  • 期刊论文

  • 发表日期：

    2026年4月4日

查看全文
  • M. A. Gul,
  • A. Kanwal,
  • Xiaoyong Yang,
  • R. S. Awan,
  • I. Khan,
  • Y. Hu,
  • M. Faisal

摘要

Galena, a primary lead ore, is found in significant deposits globally, including sediment-hosted lead–zinc, volcanogenic massive sulfide, porphyry copper, skarn, and epithermal gold deposits, often associated with valuable metals such as Ag, Sb, Tl, and Sn. The classification of galena based on trace elements is crucial for understanding metallogenic origins and improving mineral exploration strategies. This study utilized machine learning algorithms, including random forest (RF), gradient boosting (GB), multilayer perceptron (MLP), and support vector machine (SVM), to classify galena samples from 37 Pb–Zn deposits worldwide. To address dataset imbalance, resampling techniques including the synthetic minority oversampling technique (SMOTE) and random undersampling with clustering (RUC) were employed to enhance model performance and reduce bias. K-fold cross-validation (k = 10) was implemented to ensure model robustness. RF and GB achieved the highest accuracy, 97.62% and 98.19%, respectively, on oversampled data, while MLP and SVM demonstrated lower sensitivity to resampling, with SVM performing least effectively on undersampled data (86.65% accuracy). Key trace elements, such as Sn, Tl, and Ag, were identified as significant discriminators across deposit types, offering insights into ore-forming processes. Additionally, metamorphic recrystallization in galena samples, as indicated by trace element redistribution under varying geochemical conditions, provides further metallogenic insights. The use of t-distributed stochastic neighbor embedding (t-SNE) revealed distinct clustering patterns among galena samples, validating trace element data for metallogenic classification. These findings have significant implications for mineral exploration, particularly in data-limited regions, as the integration of machine learning with trace element geochemistry provides an efficient tool for targeting Pb–Zn deposits and identifying metamorphic overprinting zones, thereby optimizing exploration efforts.

Graphical Abstract