Discriminating Deposit Types Using Chlorite Trace Elements Based on Machine Learning
-
摘要: 为了解绿泥石微量元素对不同成因矿床类型是否能够进行有效的分类判别，收集了13个来自斑岩型、矽卡岩型和浅成低温热液型3种类型矿床中的2 928条绿泥石微量元素数据，采用随机森林、支持向量机和人工神经网络3种不同的机器学习算法对矿床成因类型建立了分类模型并进行特征重要性分析.结果表明，依据Ni、Cr、Co、Sr、V、Zn 6种微量元素所建立的支持向量机模型分类效果最优，其Kappa系数最高为0.89，准确率、召回率和F1值的加权平均得分为0.96，Ni、V、Co为最关键的3个判别元素.基于绿泥石矿物微量元素结合机器学习方法能够实现对矿床类型的判别，对于区域尺度找矿勘查快速评价具有重要的指示意义.Abstract: In order to study whether chlorite trace elements can effectively classify different genetic deposit types, in this paper, 2 928 trace element data of chlorite from 13 different deposits were collected, which belong to three distinct genetic types, including porphyry, skarn and epithermal deposits.Three different machine learning algorithms, including random forest, support vector machine (SVM) and artificial neural network, were used to establish classification models for the genetic types of deposits and analyze the importance of characteristics.The results show that the SVM model based on Ni, Cr, Co, Sr, V and Zn, 6 trace elements have the best classification effect, the highest Kappa coefficient is 0.89, the weighted average score of precision, recall and F1 value reach 0.96, and Ni, V and Co are the three most critical discriminant elements. In this paper it fully confirms that the machine learning classification model based on chlorite mineral trace elements can discriminate deposit types and provide an important indicator for the rapid evaluation of regional scale prospecting.
-
Key words:
- chlorite /
- trace element /
- genetic type of deposit /
- machine learning /
- mineral deposits
-
图 1 本研究使用的斑岩型、矽卡岩型和浅成低温热液型矿床中的绿泥石样本分布位置
Fig. 1. Distribution locations of chlorite samples from porphyry, skarn and epithermal deposits used in this study
图 2 绿泥石微量元素原始分布直方图
Fig. 2. Original distribution histograms of trace elements in chlorite
图 3 绿泥石微量元素对数变换后的分布直方图
Fig. 3. Distribution histograms of trace elements in chlorite after logarithm transformation
图 4 绿泥石微量元素之间的Spearman相关系数(系数值越大颜色越深)
Fig. 4. Spearman correlation coefficient among trace elements of chlorite (the higher the coefficient value, the darker the color)
图 5 模型的学习曲线支持向量机(SVM)(a)，随机森林(RF)(b)和人工神经网络(ANN)(c)
Fig. 5. Learning curve of the support vector machine (a), random forest (b) and artificial neural network (c)
图 12 SVM模型中每个类别的特征重要性
a.斑岩类型; b.浅成低温热液类型; c.矽卡岩类型；横坐标为每个特征对每个样本数据的平均SHAP值大小，蓝色的点表示元素浓度低的样本，红色代表元素浓度高的样本
Fig. 12. Feature importance of each category in SVM model
图 13 RF模型中每个类别的特征重要性
a.斑岩类型；b.浅成低温热液类型；c.矽卡岩类型
Fig. 13. Feature importance of each category in RF model
图 14 ANN模型中每个类别的特征重要性
a.斑岩类型；b.浅成低温热液类型；c.矽卡岩类型
Fig. 14. Feature importance of each category in ANN model
表 1 不同来源绿泥石的矿床名称、地理位置、成因类型及预处理前后样本数量
Table 1. The deposit, geographical locations, genetic types of chlorite from different sources and the number of samples before and after pretreatment
矿床名称 经纬度(°)
(纬度，经度)矿床类型 数据量 处理后数据量 参考文献 1 土乌铜矿 (42，92) 斑岩型 147 145 Xiao et al.(2018a) 2 土乌延东铜矿 (42，93) 斑岩型 50 50 Xiao et al.(2018b) 3 Northparkes铜金矿 (-32，148) 斑岩型 924 383 Pacey et al.(2020) 4 Resolution铜钼矿 (33，-112) 斑岩型 82 59 Cooke et al.(2020) 5 Batu Hijau铜金矿 (-8，118) 斑岩型 97 29 Wilkinson et al.(2015) 6 迪彦钦阿木钼矿 (46，118) 斑岩型 16 16 王帏等(2021) 7 小柯勒河铜钼矿 (55，126) 斑岩型 510 509 Xiao and Chen(2020)；Feng et al.(2022) 8 El Teniente铜钼矿 (-34，-71) 斑岩型 651 269 Wilkinson et al.(2020) 9 沙溪铜金矿 (31，117) 斑岩型 23 23 何光辉等(2018) 10 都龙锡锌多金属矿 (22，104) 矽卡岩型 55 53 刘仕玉等(2022) 11 铜绿山铜金铁矿 (30，115) 矽卡岩型 67 66 Zhang et al.(2020) 12 牛圈银金铅锌矿 (41，116) 浅成低温热液型 127 127 Wang et al.(2018) 13 落布岗木 (29，87) 浅成低温热液型 179 179 未刊资料 
下载: 导出CSV
表 4 模型的最优超参数
Table 4. Optimal hyperparameters of models
算法 交叉验证得分 超参数取值 随机森林 0.95 决策树数目取110；决策树最大深度取13；决策树最优模型时考虑的最大特征数取5 支持向量机 0.95 核函数取非线性核函数径向基函数(RBF)；惩罚系数取10；核函数的超参数取0.3 人工神经网络 0.94 隐藏层取2层；第一层节点数取190；第二层节点数取50；学习率取0.001；激活函数取tanh；损失优化器取adam
下载: 导出CSV
表 5 支持向量机(SVM)模型的精度评价指标
Table 5. Evaluation indexes of the support vector machine model
矿床类型 准确率 召回率 F1值 支持度 斑岩型 0.97 0.98 0.98 445 浅成低温热液型 0.94 0.92 0.93 92 矽卡岩型 0.87 0.75 0.81 36 加权平均得分 0.96 0.96 0.96 573 Kappa系数 0.89
下载: 导出CSV
表 6 随机森林(RF)模型的精度评价指标
Table 6. Evaluation indexes of the random forest model
矿床类型 准确率 召回率 F1值 支持度 斑岩型 0.96 0.99 0.97 445 浅成低温热液型 0.94 0.95 0.94 92 矽卡岩型 0.96 0.64 0.77 36 加权平均得分 0.96 0.96 0.96 573 Kappa系数 0.88
下载: 导出CSV
表 7 人工神经网络(ANN)模型的精度评价指标
Table 7. Evaluation indexes of the artificial neural network model
矿床类型 准确率 召回率 F1值 支持度 斑岩型 0.97 0.98 0.97 445 浅成低温热液型 0.94 0.95 0.94 92 矽卡岩型 0.86 0.69 0.77 36 加权平均得分 0.96 0.96 0.96 573 Kappa系数 0.88
下载: 导出CSV
-
Abbaszadeh, M., Hezarkhani, A., Soltani-Mohammadi, S., 2015. Classification of Alteration Zones Based on Whole-Rock Geochemical Data Using Support Vector Machine. Journal of the Geological Society of India, 85(4): 500-508. https://doi.org/10.1007/s12594-015-0242-3 Bédard, É., De Bronac de Vazelhes, V., Beaudoin, G., 2022. Performance of Predictive Supervised Classification Models of Trace Elements in Magnetite for Mineral Exploration. Journal of Geochemical Exploration, 236: 106959. https://doi.org/10.1016/j.gexplo.2022.106959 Chu, G. B., Zhang, S. T., Zhang, X. B., et al., 2020. Chlorite Chemistry of Tongshankou Porphyry-Related Cu-Mo-W Skarn Deposit, Eastern China: Implications for Hydrothermal Fluid Evolution and Exploration Vectoring to Concealed Orebodies. Ore Geology Reviews, 122: 103531. https://doi.org/10.1016/j.oregeorev.2020.103531 Cooke, D. R., Wilkinson, J. J., Baker, M., et al., 2020. Using Mineral Chemistry to Aid Exploration: A Case Study from the Resolution Porphyry Cu-Mo Deposit, Arizona. Economic Geology, 115(4): 813-840. https://doi.org/10.5382/econgeo.4735 Emmings, J. F., Poulton, S. W., Walsh, J., et al., 2022. Pyrite Mega-Analysis Reveals Modes of Anoxia through Geological Time. Science Advances, 8(11): eabj5687. https://doi.org/10.1126/sciadv.abj5687 Fan, Y., Zhang, W., Liu, Y. N., et al., 2021. Geochemical Characteristics of Chlorite in the Luohe Iron Deposit in the Middle-Lower Yangtze Metallogenic Belt, Eastern China. Ore Geology Reviews, 133: 104062. https://doi.org/10.1016/j.oregeorev.2021.104062 Feng, Y. Z., Chu, G. B., Xiao, B., et al., 2022. Chlorite Mineralogy, Geochemistry and Exploration Implications: a Case Study of the Xiaokelehe Porphyry Cu-Mo Deposit in NE China. Ore Geology Reviews, 140: 104568. https://doi.org/10.1016/j.oregeorev.2021.104568 Fu, G. M., Lü, Q. T., Yan, J. Y., et al., 2021.3D Mineral Prospectivity Modeling Based on Machine Learning: A Case Study of the Zhuxi Tungsten Deposit in Northeastern Jiangxi Province, South China. Ore Geology Reviews, 67(Suppl. 1): 275-276 (in Chinese with English abstract). Guo, P., Yang, T., Xu, W. L., et al., 2021. Machine Learning Reveals Source Compositions of Intraplate Basaltic Rocks. Geochemistry, Geophysics, Geosystems, 22(9): e2021GC009946. https://doi.org/10.1029/2021gc009946 Han, S., Li, M. C., Ren, Q. B., et al., 2018. Intelligent Determination and Data Mining for Tectonic Settings of Basalts Based on Big Data Methods. Acta Petrologica Sinica, 34(11): 3207-3216(in Chinese with English abstract). He, G. H., Zhou, T. F., Fan, Y., et al., 2018. Geochemical Characteristics and Exploration Implications of Chlorite in Shaxi Porphyry Copper Gold Deposit, Lujiang. Mineral Deposits, 37(6): 1247-1259(in Chinese with English abstract). Hong, S., Zuo, R. G., Huang, X. W., et al., 2021. Distinguishing IOCG and IOA Deposits via Random Forest Algorithm Based on Magnetite Composition. Journal of Geochemical Exploration, 230: 106859. https://doi.org/10.1016/j.gexplo.2021.106859 Huang, J. H., Chen, H. Y., Han, J. S., et al., 2018. Alteration Zonation and Short Wavelength Infrared (SWIR) Characteristics of the Honghai VMS Cu-Zn Deposit, Eastern Tianshan, NW China. Ore Geology Reviews, 100: 263-279. https://doi.org/10.1016/j.oregeorev.2017.02.037 Jiao, S. T., Zhou, Y. Z., Zhang, Q., et al., 2018. Study on Intelligent Discrimination of Tectonic Settings Based on Global Gabbro Data from GEOROC. Acta Petrologica Sinica, 34(11): 3189-3194(in Chinese with English abstract). Li, C. B., Xiao, K. Y., Li, N., et al., 2020. A Comparative Study of Support Vector Machine, Random Forest and Artificial Neural Network Machine Learning Algorithms in Geochemical Anomaly Information Extraction. Acta Geoscientica Sinica, 41(2): 309-319(in Chinese with English abstract). Li, C. H., Shen, P., Zhao, Y., et al., 2022. Mineral Chemistry of Chlorite in Different Geologic Environments and Its Implications for Porphyry Cu±Au±Mo Deposits. Ore Geology Reviews, 149: 105112. https://doi.org/10.1016/j.oregeorev.2022.105112 Li, S., Chen, J. P., Liu, C., et al., 2021. Mineral Prospectivity Prediction via Convolutional Neural Networks Based on Geological Big Data. Journal of Earth Science, 32(2): 327-347. https://doi.org/10.1007/s12583-020-1365-z Liu, B., He, J. R., Geng, Y. J., et al., 2017. Recent Advances in Infrastructure Architecture of Parallel Machine Learning Algorithms. Computer Engineering and Applications, 53(11): 31-38 (in Chinese with English abstract). Liu, L., Shen, J. K., Zhang, L. X., et al., 2023. A Machine Learning‐Based Method for Rapid Prediction of Earthquake Damage in Brick Masonry Houses. Earth Science, 48(5): 1769-1779(in Chinese with English abstract). Liu, S. Y., Liu, Y. P., Ye, L., et al., 2022. Genetic Mineralogy of Chlorite in the Dulong Sn-Zn Polymetallic Deposit in Maguan, Yunnan Province, China. Acta Mineralogica Sinica, 42(1): 1-13 (in Chinese with English abstract). Liu, Y. P., Zhu, L. X., Zhou, Y. Z., 2020. Experimental Research on Big Data Mining and Intelligent Prediction of Prospecting Target Area: Application of Convolutional Neural Network Model. Geotectonica et Metallogenia, 44(2): 192-202(in Chinese with English abstract). Mao, M., Rukhlov, A. S., Rowins, S. M., et al., 2016. Apatite Trace Element Compositions: A Robust New Tool for Mineral Exploration. Economic Geology, 111(5): 1187-1222. https://doi.org/10.2113/econgeo.111.5.1187 Martínez-Serrano, R. G., Dubois, M., 1998. Chemical Variations in Chlorite at the Los Humeros Geothermal System, Mexico. Clays and Clay Minerals, 46(6): 615-628. https://doi.org/10.1346/CCMN.1998.0460602 Nathwani, C. L., Wilkinson, J. J., Fry, G., et al., 2022. Machine Learning for Geochemical Exploration: Classifying Metallogenic Fertility in Arc Magmas and Insights into Porphyry Copper Deposit Formation. Mineralium Deposita, 57(7): 1143-1166. https://doi.org/10.1007/s00126-021-01086-9 Pacey, A., Wilkinson, J. J., Cooke, D. R., 2020. Chlorite and Epidote Mineral Chemistry in Porphyry Ore Systems: A Case Study of the Northparkes District, New South Wales, Australia. Economic Geology, 115(4): 701-727. https://doi.org/10.5382/econgeo.4700 Petrelli, M., Perugini, D., 2016. Solving Petrological Problems through Machine Learning: The Study Case of Tectonic Discrimination Using Geochemical and Isotopic Data. Contributions to Mineralogy and Petrology, 171(10): 81. https://doi.org/10.1007/s00410-016-1292-2 Rodriguez-Galiano, V., Sanchez-Castillo, M., Chica-Olmo, M., et al., 2015. Machine Learning Predictive Models for Mineral Prospectivity: An Evaluation of Neural Networks, Random Forest, Regression Trees and Support Vector Machines. Ore Geology Reviews, 71: 804-818. https://doi.org/10.1016/j.oregeorev.2015.01.001 Rudin, C., 2019. Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead. Nature Machine Intelligence, 1(5): 206-215. https://doi.org/10.1038/s42256-019-0048-x Wang, W., Leng, C. B., Zhang, X. C., 2021. Mineral Chemistry of Chlorite and Epidote in Diyanqinamu Porphyry Mo Deposit of Inner Mongolia and Its Implication for Exploration. Mineral Deposits, 40(2): 241-261(in Chinese with English abstract). Wang, Y., Qiu, K. F., Müller, A., et al., 2021. Machine Learning Prediction of Quartz Forming-Environments. Journal of Geophysical Research: Solid Earth, 126(8): e2021JB021925. https://doi.org/10.1029/2021jb021925 Wang, Y. J., Fan, H. H., Pang, Y. Q., et al., 2022. Geochemical Characteristics of Chlorite in Xiangshan Uranium Ore Field, South China and Its Exploration Implication. Minerals, 12(6): 693. https://doi.org/10.3390/min12060693 Wang, Z. Q., Chen, B., Yan, X., et al., 2018. Characteristics of Hydrothermal Chlorite from the Niujuan Ag-Au-Pb-Zn Deposit in the North Margin of NCC and Implications for Exploration Tools for Ore Deposits. Ore Geology Reviews, 101: 398-412. https://doi.org/10.1016/j.oregeorev.2018.08.003 Wilkinson, J. J., Baker, M. J., Cooke, D. R., et al., 2020. Exploration Targeting in Porphyry Cu Systems Using Propylitic Mineral Chemistry: A Case Study of the El Teniente Deposit, Chile. Economic Geology, 115(4): 771-791. https://doi.org/10.5382/econgeo.4738 Wilkinson, J. J., Chang, Z. S., Cooke, D. R., et al., 2015. The Chlorite Proximitor: A New Tool for Detecting Porphyry Ore Deposits. Journal of Geochemical Exploration, 152: 10-26. https://doi.org/10.1016/j.gexplo.2015.01.005 Xiao, B., Chen, H. Y., 2020. Elemental Behavior during Chlorite Alteration: New Insights from a Combined EMPA and LA-ICPMS Study in Porphyry Cu Systems. Chemical Geology, 543: 119604. https://doi.org/10.1016/j.chemgeo.2020.119604 Xiao, B., Chen, H. Y., Wang, Y. F., et al., 2018a. Chlorite and Epidote Chemistry of the Yandong Cu Deposit, NW China: Metallogenic and Exploration Implications for Paleozoic Porphyry Cu Systems in the Eastern Tianshan. Ore Geology Reviews, 100: 168-182. https://doi.org/10.1016/j.oregeorev.2017.03.004 Xiao, B., Chen, H. Y., Hollings, P., et al., 2018b. Element Transport and Enrichment during Propylitic Alteration in Paleozoic Porphyry Cu Mineralization Systems: Insights from Chlorite Chemistry. Ore Geology Reviews, 102: 437-448. https://doi.org/10.1016/j.oregeorev.2018.09.020 Yang, L., Wei, J., 2023. Prediction of Rockburst Intensity Grade Based on SVM and Adaptive Boosting Algorithm. Earth Science, 48(5): 2011-2023(in Chinese with English abstract). Zhang, S. T., Xiao, B., Long, X. P., et al., 2020. Chlorite as an Exploration Indicator for Concealed Skarn Mineralization: Perspective from the Tonglushan Cu-Au-Fe Skarn Deposit, Eastern China. Ore Geology Reviews, 126: 103778. https://doi.org/10.1016/j.oregeorev.2020.103778 Zhou, Y. Z., Zhang, Z. J., Yang, J., et al., 2022. Machine Learning and Singularity Analysis Reveal Zircon Fertility and Magmatic Intensity: Implications for Porphyry Copper Potential. Natural Resources Research, 31(6): 3061-3078. https://doi.org/10.1007/s11053-022-10122-y 付光明, 严加永, 罗凡, 等, 2021. 基于随机森林算法的三维成矿预测: 以赣东北朱溪钨矿床外围为例. 地质论评, 67(增刊1): 275-276. 韩帅, 李明超, 任秋兵, 等, 2018. 基于大数据方法的玄武岩大地构造环境智能挖掘判别与分析. 岩石学报, 34(11): 3207-3216. 何光辉, 周涛发, 范裕, 等, 2018. 庐江沙溪斑岩型铜金矿床绿泥石的地球化学特征及找矿指示. 矿床地质, 37(6): 1247-1259. 焦守涛, 周永章, 张旗, 等, 2018. 基于GEOROC数据库的全球辉长岩大数据的大地构造环境智能判别研究. 岩石学报, 34(11): 3189-3194. 李苍柏, 肖克炎, 李楠, 等, 2020. 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. 地球学报, 41(2): 309-319. 刘斌, 何进荣, 耿耀君, 等, 2017. 并行机器学习算法基础体系前沿进展综述. 计算机工程与应用, 53(11): 31-38, 89. 刘丽, 沈俊凯, 张令心, 2023. 基于机器学习的砖砌体房屋震害快速预测方法. 地球科学, 48(5): 1769-1779. 刘仕玉, 刘玉平, 叶霖, 等, 2022. 云南马关都龙锡锌多金属矿床绿泥石成因矿物学. 矿物学报, 42(1): 1-13. 刘艳鹏, 朱立新, 周永章, 2020. 大数据挖掘与智能预测找矿靶区实验研究: 卷积神经网络模型的应用. 大地构造与成矿学, 44(2): 192-202. 王帏, 冷成彪, 张兴春, 等, 2021. 内蒙古迪彦钦阿木斑岩钼矿床绿泥石和绿帘石矿物化学特征及其成矿指示意义. 矿床地质, 40(2): 241-261. 杨玲, 魏静, 2023. 基于支持向量机和增强学习算法的岩爆烈度等级预测. 地球科学, 48(5): 2011-2023. doi: 10.3799/dqkx.2022.251 -
dqkxzx-49-12-4303-附表1.xlsx
-
点击查看大图
计量
- 文章访问数: 442
- HTML全文浏览量: 261
- PDF下载量: 105
- 被引次数: 0
