| 支持向量机、随机森林和人工神经网络机器学习算法在地球化学异常信息提取中的对比研究 |
| |
| 引用本文： | 李苍柏,肖克炎,李楠,宋相龙,张帅,王凯,楚文楷,曹瑞.支持向量机、随机森林和人工神经网络机器学习算法在地球化学异常信息提取中的对比研究[J].地球学报,2020,41(2):309-319. |
| |
| 作者姓名： | 李苍柏 肖克炎 李楠 宋相龙 张帅 王凯 楚文楷 曹瑞 |
| |
| 作者单位： | 中国地质科学院矿产资源研究所, 自然资源部成矿作用与资源评价重点实验室;中国地质大学(北京),中国地质科学院矿产资源研究所, 自然资源部成矿作用与资源评价重点实验室,中国地质科学院矿产资源研究所, 自然资源部成矿作用与资源评价重点实验室,中国地质科学院,中国地质科学院矿产资源研究所, 自然资源部成矿作用与资源评价重点实验室;中国地质大学(北京),中国地质科学院矿产资源研究所, 自然资源部成矿作用与资源评价重点实验室;中国地质大学(北京),中国地质科学院矿产资源研究所, 自然资源部成矿作用与资源评价重点实验室;中国地质大学(北京),中国地质科学院矿产资源研究所, 自然资源部成矿作用与资源评价重点实验室;中国地质大学(北京) |
| |
| 基金项目： | 国家自然科学基金面上项目“基于地质先验模型的区域大比例尺三维地质建模关键技术研究”(编号: 41672330);国家重点研发计划“深地资源勘查开采”重点专项课题“深部矿产三维可视化预测评价软件系统研发”(编号: 2017YFC0601501);中国地质调查局地质调查项目“全国矿产资源潜力动态评价”(编号: DD20190193) |
| |
| 摘 要： | 地球化学勘查是通过发现异常、解释评价异常进行找矿的。因此,地球化学异常识别对矿产资源的定位、定量预测具有重要的的指示作用。在大数据时代的背景下,机器学习方法不要求数据满足正态分布的分布形式,且具有非线性以及泛化能力强等特点,因而逐渐地被应用于矿产资源的定量预测评价,如神经网络、支持向量机、贝叶斯网络、随机森林、受限玻尔兹曼机、极限学习机等。本文通过设计理论实验,可视化了不同算法,提出了不同机器学习方法在不同地区的地球化学异常信息提取中的效果存在不一致性的假设。在此基础上,以湖南香花岭锡多金属矿整装勘查区及甘肃合作金矿整装勘查区的地球化学异常提取为研究内容,将人工神经网络、随机森林以及支持向量机应用于研究区地球化学异常信息的提取与识别工作。在香花岭研究区,人工神经网络的结果较好,在合作研究区,随机森林的结果较好,从而验证了上述假设。通过生成两研究区的地球化学异常图,讨论了该方法在两研究区地球化学异常的地质意义和该方法的可靠性与实用性。此外,还完善了基于多种监督机器学习方法的地球化学异常信息提取流程,为软件开发提供了一定的理论依据。
|
| 关 键 词： | 机器学习 地球化学异常 人工神经网络 随机森林 支持向量机 |
A Comparative Study of Support Vector Machine, Random Forest and Artificial Neural Network Machine Learning Algorithms in Geochemical Anomaly Information Extraction |
| |
| LI Cang-bai,XIAO Ke-yan,LI Nan,SONG Xiang-long,ZHANG Shuai,WANG Kai,CHU Wen-kai and CAO Rui.A Comparative Study of Support Vector Machine, Random Forest and Artificial Neural Network Machine Learning Algorithms in Geochemical Anomaly Information Extraction[J].Acta Geoscientia Sinica,2020,41(2):309-319. |
| |
| Authors: | LI Cang-bai XIAO Ke-yan LI Nan SONG Xiang-long ZHANG Shuai WANG Kai CHU Wen-kai and CAO Rui |
| |
| Institution: | MNR Key Laboratory of Metallogeny and Mineral Assessment, Institute of Mineral Resources,Chinese Academy of Geological Sciences;China University of Geosciences (Beijing),MNR Key Laboratory of Metallogeny and Mineral Assessment, Institute of Mineral Resources,Chinese Academy of Geological Sciences,MNR Key Laboratory of Metallogeny and Mineral Assessment, Institute of Mineral Resources,Chinese Academy of Geological Sciences,Chinese Academy of Geological Sciences,MNR Key Laboratory of Metallogeny and Mineral Assessment, Institute of Mineral Resources,Chinese Academy of Geological Sciences;China University of Geosciences (Beijing),MNR Key Laboratory of Metallogeny and Mineral Assessment, Institute of Mineral Resources,Chinese Academy of Geological Sciences;China University of Geosciences (Beijing),MNR Key Laboratory of Metallogeny and Mineral Assessment, Institute of Mineral Resources,Chinese Academy of Geological Sciences;China University of Geosciences (Beijing) and MNR Key Laboratory of Metallogeny and Mineral Assessment, Institute of Mineral Resources,Chinese Academy of Geological Sciences;China University of Geosciences (Beijing) |
| |
| Abstract: | Geochemical prospecting is conducted through finding anomalies as well as explaining and evaluating anomalies.Therefore,geochemical anomaly identification has a certain indicative effect on the locating and quantitative predicting of mineral resources.In the background of big data era,the machine learning method doesn’t need the distribution form that obeys the normal distribution,and has the characteristics of nonlinear and strong generalization.Thus,it is gradually applied to the quantitative prediction of mineral resources,such as Neural Network,Support Vector Machine,Bayesian Network,Random Forest,Restricted Boltzmann Machine,and Extreme Learning Machine.Based on previous studies,this paper proposes the hypothesis that the effects of different supervised machine learning methods on geochemical anomaly information extraction in different regions are inconsistent.Through the design theory experiment,the results of different algorithms in different data distributions aren’t consistent.On such a basis,the geochemical anomaly detection of the Xianghualing tin polymetallic ore deposit in Hunan Province and the Hezuo gold ore deposit in Gansu Province were taken as examples,and the resulting geochemical anomalies in the two regions were extracted and identified by Artificial Neural Network,Random Forest and Support Vector Machine.In the study area of Xianghualing,the results of Artificial Neural Network were relatively good,while in the study area of Hezuo,the results of Random Forest were relatively good,which verified the above hypothesis.The geological significance of the method and its reliability and practicability were discussed by generating geochemical anomaly maps in the two study areas.In addition,the geochemical anomaly information extraction process based on a variety of supervised machine learning methods was also improved,which provides a certain theoretical basis for software development. |
| |
| Keywords: | machine learning geochemical anomalies Artificial Neural Network Random Forest Support Vector Machine |
| 本文献已被 维普 等数据库收录！ |
| 点击此处可从《地球学报》浏览原始摘要信息 |
| 点击此处可从《地球学报》下载免费的PDF全文 |
|
