地质通报2026，Vol.45Issue(8)：1483-1502,20.DOI:10.12097/gbc.2025.03.005
机器学习驱动矿床成因类型判别
Machine learning driven discrimination of ore genesis types:A case study of the Hua'aobaote Pb-Zn-Ag deposit in the southern Great Xing'an Range
摘要
Abstract
[Objective]Accurately identifying the genesis of mineral deposits is a crucial step in mineral exploration.Traditional methods,relying on expert experience and limited data,are often constrained by inefficiencies and accuracy limitations.The trace elements in sphalerite can precisely reflect the ore-forming environment,serving as key indicators for distinguishing the genesis of Pb-Zn deposits.To improve the efficiency and accuracy of genesis classification,exploring the application of low-code machine learning methods in geological data analysis is essential.[Methods]This study utilizes the PyCaret machine learning library in combination with trace element data from sphalerite to build a classification model for deposit genesis,covering porphyry,skarn,and hydrothermal vein types.The model is applied to the Hua'aobaote Pb-Zn-Ag deposit in the Great Xing'an Range.[Results]After comparing multiple models,the ET and LightGBM models performed the best,achieving accuracies of 97.59%and 96.88%,respectively.Feature importance analysis revealed that Co and Fe elements play a crucial role in distinguishing deposit genesis and are significant factors influencing model decisions.[Conclusions]Based on the model results and previous studies,the Hua'aobaote deposit is classified as a low-temperature,shallow hydrothermal vein deposit associated with a porphyry-style magmatic-hydrothermal system.This study demonstrates the feasibility and practicality of low-code machine learning methods for ore deposit genesis analysis,offering an effective technological approach for rapid and accurate genesis classification.关键词
机器学习/闪锌矿微量元素/矿床成因/花敖包特/大兴安岭南段/矿产勘查工程Key words
machine learning/trace elements of sphalerite/deposit genesis/Hua'aobaote/southern Great Xing'an Range/mineral exploration engineering分类
天文与地球科学引用本文复制引用
王智宇,段雪鹏,宓奎峰,王达,张晓暄..机器学习驱动矿床成因类型判别[J].地质通报,2026,45(8):1483-1502,20.基金项目
深地国家科技重大专项《大兴安岭中南段致矿异常信息融合及多尺度找矿指标体系构建》(编号:2024ZD1001506)、中国地质科学院基本科研业务费《基于锆石微量元素和机器学习的大兴安岭南段花岗岩体成矿潜力判别》(编号:JKYQN202402)和中国地质调查局项目《重要锡锌等战略性矿产选区调查评价》(编号:DD202602109804) Supported by the National Major Science and Technology Projects of China(No.2024ZD1001506),the Basic Research Fund of the Chinese Academy of Geological Sciences(No.JKYQN202402),and the project of China Geological Survey(No.DD202602109804) （编号:2024ZD1001506）
