Scholarlink

<p>With increasing need for domestically sourced critical metals in the European Union (EU), particularly rare earth elements (REEs), innovative methods are needed to identify areas favorable to mineralization of these critical metals. Over the past 20&#xa0;years, numerous predictive algorithms, including machine learning (ML)-based methods have been applied to identify areas of strong mineral potential. The EU-funded “Exploration Information Systems” project has created an open-source GIS-based toolkit for mineral potential mapping including several data-driven ML-based algorithms. In this study, we employed logistic regression (LR) and random forest (RF) classification to test this toolkit to identify areas with potential for Bastnäs-type REE mineralization in the REE-Line in the Bergslagen district of south-central Sweden. Using a mineral systems model to translate key formational factors into mappable proxies, predictive maps were generated using both algorithms. Models were evaluated using confusion matrices and receiver operator characteristic curves. The results showed that RF performed better than LR. The former returned robust results for most training and test sets, with average accuracy of 87%, and identified 20% of the total area of the REE-Line as prospective, capturing over 90% of known deposits. Stakeholder validation using issued exploration claim areas showed many of the areas identified as being prospective are currently being explored for the REE, demonstrating the potential of RF using a mineral system model for REE 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

Random Forest Modeling of Bastnäs-Type REE Deposits in Bergslagen, Central Sweden: Testing the EIS Toolkit for Mineral Potential Mapping

  • 期刊论文

  • 发表日期：

    2026年3月15日

查看全文
  • Patrick Casey,
  • Martiya Sadeghi

摘要

With increasing need for domestically sourced critical metals in the European Union (EU), particularly rare earth elements (REEs), innovative methods are needed to identify areas favorable to mineralization of these critical metals. Over the past 20 years, numerous predictive algorithms, including machine learning (ML)-based methods have been applied to identify areas of strong mineral potential. The EU-funded “Exploration Information Systems” project has created an open-source GIS-based toolkit for mineral potential mapping including several data-driven ML-based algorithms. In this study, we employed logistic regression (LR) and random forest (RF) classification to test this toolkit to identify areas with potential for Bastnäs-type REE mineralization in the REE-Line in the Bergslagen district of south-central Sweden. Using a mineral systems model to translate key formational factors into mappable proxies, predictive maps were generated using both algorithms. Models were evaluated using confusion matrices and receiver operator characteristic curves. The results showed that RF performed better than LR. The former returned robust results for most training and test sets, with average accuracy of 87%, and identified 20% of the total area of the REE-Line as prospective, capturing over 90% of known deposits. Stakeholder validation using issued exploration claim areas showed many of the areas identified as being prospective are currently being explored for the REE, demonstrating the potential of RF using a mineral system model for REE exploration.