机器学习方法在矿产资源定量预测应用研究进展  

Advances in the application of machine learning methods in mineral prospectivity mapping

作　　者：马瑶 赵江南[1] Ma Yao;Zhao Jiangnan(School of Earth Resources,China University of Geosciences(Wuhan),Wuhan 430074,China)

机构地区：[1]中国地质大学(武汉)资源学院,武汉430074

出　　处：《地质科技通报》2021年第1期132-141,共10页Bulletin of Geological Science and Technology

基　　金：国家重点研发计划(2016YFC0600509)。

摘　　要：回顾了国内外在矿产资源定量预测研究领域的发展历程,对近十年来国外相关方向的文献进行了统计对比分析,结果显示机器学习方法已经成为矿产资源定量预测研究领域的热点方向,并主要在如下3个方面发挥了积极的作用:①提取和挖掘复杂数据中隐藏的难以识别的矿化信息;②致矿异常信息关联与转换;③多源地学数据的致矿异常信息融合、预测和发现矿床。对逻辑回归、人工神经网络、随机森林与支持向量机等主要机器学习算法与模型在矿产资源定量预测实践中的应用效果进行了评述,并探讨了在实际应用过程中存在的样本选择、错分代价、不确定性评价以及模型性能评价等主要问题及目前的解决方案。最后提出基于大数据与机器学习的矿产资源定量预测是未来发展的重要趋势。This paper reviews the development of mineral prospectivity mapping at home and abroad,and conducts statistical comparative analysis of relevant foreign literature in the past decade.It shows that machine learning methods have become a hot topic in the field of mineral prospectivity mapping,and have played an active role in the following three aspects:①extraction and mining of hidden and unrecognizable mineralization information in complex data;②association and transformation of ore-forming anomaly information;③fusion,prediction and discovery of ore-forming anomaly information from multi-source geological data.Firstly,the application effects of major machine learning algorithms and models,such as logistic regression,artificial neural networks,random forests,and support vector machines,in mineral prospectivity mapping are reviewed.Secondly,it discusses the main problems in the application process,such as sample selection,misclassification cost,uncertainty evaluation,and model performance evaluation,as well as the current solutions.Finally,it is proposed that quantitative prediction of mineral resources based on big data and machine learning is an important trend in the future.

关 键 词：矿产资源定量预测 地学大数据 数据挖掘 信息融合 机器学习 

分 类 号：P624.7[天文地球—地质矿产勘探]