Scholarlink

<p>The Tobique-Chaleur Zone (TCZ), located in northern New Brunswick and the neighboring Gaspé Peninsula, hosts a number of gold occurrences. Gold mineralization in TCZ is spatially associated with large-scale crustal structures or their subsidiary splays (e.g., the Rocky Brook-Millstream Fault system). The gold occurrences in the Chaleur Bay Synclinorium, are hosted primarily underlain by Upper Silurian to Early Devonian strata, with some lower Silurian strata also present. This study applies three supervised machine learning algorithms Random Forest (RF), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGBoost) to generate mineral prospectivity mapping (MPM) for epithermal gold mineralization in the Tobique–Chaleur Zone of northern New Brunswick. A total of 24 evidence layers, derived from geological, geochemical, and geophysical datasets, were selected based on a mineral systems framework to represent key ore-forming processes including source, pathway, and trap components. Model performance was evaluated using Receiver Operating Characteristic (ROC) curves, with RF achieving the highest AUC (0.93), followed by SVM (0.92) and XGBoost (0.91). Additionally, prediction–area (P–A) plot analysis revealed that XGBoost was the most efficient model in terms of spatial targeting. A majority voting ensemble and confidence index map were employed to enhance robustness and quantify model uncertainty. The spatial correlation of high-prospectivity zones with known gold occurrences and key geological features validates the applied approach. These results demonstrate the effectiveness of integrated machine learning and mineral systems modeling in generating reliable and interpretable prospectivity maps for mineral exploration.</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

Machine learning-based mineral prospectivity mapping of epithermal gold mineralization in Northern new brunswick: a comparative study of random forest, support vector machine, and XGboost classifiers

  • 期刊论文

  • 发表日期：

    2025年11月17日

查看全文
  • Farzaneh Mami Khalifani,
  • David R. Lentz,
  • James A. Walker

摘要

The Tobique-Chaleur Zone (TCZ), located in northern New Brunswick and the neighboring Gaspé Peninsula, hosts a number of gold occurrences. Gold mineralization in TCZ is spatially associated with large-scale crustal structures or their subsidiary splays (e.g., the Rocky Brook-Millstream Fault system). The gold occurrences in the Chaleur Bay Synclinorium, are hosted primarily underlain by Upper Silurian to Early Devonian strata, with some lower Silurian strata also present. This study applies three supervised machine learning algorithms Random Forest (RF), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGBoost) to generate mineral prospectivity mapping (MPM) for epithermal gold mineralization in the Tobique–Chaleur Zone of northern New Brunswick. A total of 24 evidence layers, derived from geological, geochemical, and geophysical datasets, were selected based on a mineral systems framework to represent key ore-forming processes including source, pathway, and trap components. Model performance was evaluated using Receiver Operating Characteristic (ROC) curves, with RF achieving the highest AUC (0.93), followed by SVM (0.92) and XGBoost (0.91). Additionally, prediction–area (P–A) plot analysis revealed that XGBoost was the most efficient model in terms of spatial targeting. A majority voting ensemble and confidence index map were employed to enhance robustness and quantify model uncertainty. The spatial correlation of high-prospectivity zones with known gold occurrences and key geological features validates the applied approach. These results demonstrate the effectiveness of integrated machine learning and mineral systems modeling in generating reliable and interpretable prospectivity maps for mineral exploration.