Scholarlink

<p>In recent years, three-dimensional mineral prospectivity mapping (3D MPM) has become a key approach for predicting mineral deposits at depth. Many researchers have applied machine learning (ML) methods for 3D mineralization prediction. However, the integration of ML with heterogeneous data types often leads to an increased range of predictions and increased uncertainty. In this paper, we propose an ensemble gradient boosting approach that is adaptable to different data representations, and we evaluate its impact on the resulting prediction accuracy. This method was applied in the analysis of prospectivity for Haopinggou gold polymetallic deposit in the Xiong'ershan region of the western Henan metallogenic belt, where it holds potential for deep prediction. A 3D geological model of the deposit was established from primary geological data, and a quantitative 3D prediction model was developed based on mineral system theory. The influences of different data types on the prediction outcomes were assessed using the SHapley Additive exPlanations framework. The results revealed that the gradient boosting model developed based on binary data was highly effective for 3D MPM, yielding higher area under the curve and prediction accuracy than those of the competing methods while precisely delineating deep prospecting target zones. These findings suggest high-priority aspects for guiding future drilling and deep exploration processes in the Haopinggou mining area.</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

Three-Dimensional Mineral Prospectivity Mapping by a Gradient Boosting-Based Integrated Learning Method with Data Representation Adaptability: A Case Study of the Haopinggou Gold Polymetallic Deposit, China

  • 期刊论文

  • 发表日期：

    2026年5月5日

查看全文
  • Mingjing Fan,
  • Keyan Xiao,
  • Li Sun,
  • Nan Li,
  • Yang Xu,
  • Wenjie Li

摘要

In recent years, three-dimensional mineral prospectivity mapping (3D MPM) has become a key approach for predicting mineral deposits at depth. Many researchers have applied machine learning (ML) methods for 3D mineralization prediction. However, the integration of ML with heterogeneous data types often leads to an increased range of predictions and increased uncertainty. In this paper, we propose an ensemble gradient boosting approach that is adaptable to different data representations, and we evaluate its impact on the resulting prediction accuracy. This method was applied in the analysis of prospectivity for Haopinggou gold polymetallic deposit in the Xiong'ershan region of the western Henan metallogenic belt, where it holds potential for deep prediction. A 3D geological model of the deposit was established from primary geological data, and a quantitative 3D prediction model was developed based on mineral system theory. The influences of different data types on the prediction outcomes were assessed using the SHapley Additive exPlanations framework. The results revealed that the gradient boosting model developed based on binary data was highly effective for 3D MPM, yielding higher area under the curve and prediction accuracy than those of the competing methods while precisely delineating deep prospecting target zones. These findings suggest high-priority aspects for guiding future drilling and deep exploration processes in the Haopinggou mining area.