The Kalatongke copper-nickel district is a significant producer of scarce mineral resources. However, with continuous exploitation, it faces a crisis of depleting reserves. Consequently, the discovery of new ore bodies in this district has become a primary objective, and deep exploration presents an opportunity to revitalize such a crisis-hit mining area. This study integrates 3D modeling and machine learning technologies to build district-scale geological and geophysical models for three-dimensional (3D) quantitative mineral resource prediction. The workflow consisted of four main steps: (1) Gravity, magnetic, electromagnetic, and seismic data were used to construct 3D models through geophysical inversion and spatial interpolation, yielding physical property models for density, magnetic susceptibility, resistivity, and seismic wave velocity. (2) Based on metallogenic principles and petrophysical characteristics, a lithological model was constructed using the K-means clustering machine learning method, while a fault model was built by integrating surface fault observations with deep constraints from electromagnetic and seismic sections. (3) The geological and geophysical models were converted into exploration predictor variables. The Bagging-based Positive-Unlabeled Learning (BPUL) algorithm was employed for 3D mineral prospectivity mapping utilizing Random Forest, Support Vector Machine, and XGBoost as base classifiers. Comparison revealed that the BPUL-XGBoost15 model delivered the best performance. This optimal model was used to generate a mineral prospectivity probability map, and initial targets were delineated using the prediction-volume (P-V) plot technique. (4) A risk-return analysis was conducted to evaluate the uncertainty of the optimal model. The initial targets were then refined based on this analysis to identify those with high potential returns and low exploration risk. Finally, these targets were classified according to their metallogenic geological context. The identified key targets are critical for guiding subsequent exploration efforts. The research framework established in this study can serve as a valuable reference for the deep exploration of similar types of mineral deposits.
随机森林(RF)属于集成学习算法中的一种[49],最初由Ho[50]在1995年提出,之后Bagging技术和基尼指数被引入对原RF算法进行改进[51-53]。Bagging技术基于Boostrap采样技术对样本进行有放回地随机抽样,并利用抽样样本构建新的训练子集对基分类器进行训练。RF算法中基分类器通常是决策树算法[54]中的分类回归数算法(Classification and Regression Tree,CART),基尼指数是CART算法进行特征选择用于分类的重要依据。可将RF、Bagging技术和CART算法之间的关系简单总结成:RF=Bagging+CART[14]。
LIX H, XUEC, CHENY H, et al. 3D convolutional neural network-based 3D mineral prospectivity modeling for targeting concealed mineralization within Chating area, middle-lower Yangtze River metallogenic belt, China[J]. Ore Geology Reviews, 2023, 157: 105444.
[14]
JIAR, LVY K, WANGG W, et al. A stacking methodology of machine learning for 3D geological modeling with geological-geophysical datasets, Laochang Sn camp, Gejiu (China)[J]. Computers & Geosciences, 2021, 151: 104754.
GAOM, WANGG W, XUY C, et al. 3D mineral exploration Cu-Zn targeting with multi-source geoscience datasets in the Weilasituo-Bairendaba district, Inner Mongolia, China[J]. Frontiers in Earth Science, 2023, 11: 1102640.
GAOM, WANGG W, CARRANZAE J M, et al. 3D Au targeting using machine learning with different sample combination and return-risk analysis in the Sanshandao-Cangshang district, Shandong Province, China[J]. Natural Resources Research, 2023, 33: 51-74.
GAOM, WANGG W, YANGW D, et al. Bagging-based positive-unlabeled data learning algorithm with base learners random forest and XGBoost for 3D exploration targeting in the Kalatongke district, Xinjiang, China[J]. Natural Resources Research, 2023, 32: 437-459.
[28]
ZHANGZ Q, WANGG W, LIUC, et al. Bagging-based positive-unlabeled learning algorithm with Bayesian hyperparameter optimization for three-dimensional mineral potential mapping[J]. Computers & Geosciences, 2021, 154: 104817.
[29]
ZHANGZ Q, WANGG W, CARRANZAE J M, et al. An integrated framework for data-driven mineral prospectivity mapping using Bagging-based positive unlabeled learning and Bayesian cost-sensitive logistic regression[J]. Natural Resources Research, 2022, 31: 3041-3060.
[30]
KANGZ, QINK Z, MAOY J, et al. The formation of a magmatic Ni-Cu sulfide deposit in mafic intrusions at the Kalatongke, NW China: insights from amphibole mineralogy and composition[J]. Lithos, 2019, 352/353: 105317.
[31]
CHENW, XUEG Q, ZHOUN N, et al. Delineating ore-forming rock using a frequency domain controlled-source electromagnetic method[J]. Ore Geology Reviews, 2019, 115: 103167.
[32]
LIC S, ZHANGM J, FUP E, et al. The Kalatongke magmatic Ni-Cu deposits in the Central Asian Orogenic Belt, NW China: product of slab window magmatism?[J]. Mineralium Deposita, 2012, 47: 51-67.
TANGD M, QINK Z, SUB X, et al. Sulfur and copper isotopic signatures of chalcopyrite at Kalatongke and Baishiquan: insights into the origin of magmatic Ni-Cu sulfide deposits[J]. Geochimica et Cosmochimica Acta, 2020, 275: 209-228.
[35]
QIANZ Z, DUANJ, LIC S, et al. Paleozoic mafic-intermediate intrusions (320-287 Ma) in the Kalatongke area, southern Altai, NW China: products of protracted magmatism in a convergent tectonic setting[J]. Journal of Asian Earth Sciences, 2017.
XIAOW J, ZHANGL C, QINK Z, et al. Paleozoic accretionary and collisional tectonics of the eastern Tianshan (China): implications for the continental growth of Central Asia[J]. American Journal of Science, 2004, 304(4): 370-395.
[40]
DUANJ, QIANZ Z, FENGY Q, et al. Compositional variations of several Early Permian magmatic sulfide deposits in the Kalatongke district, southern Altai, western China: with genetic and exploration implications[J]. Ore Geology Reviews, 2017, 90: 1-15.
TANGZ L, YANH Q, JIAOJ G, et al. Regional metallogenic controls of small-intrusion-hosted Ni-Cu (PGE) ore deposits in China[J]. Earth Science Frontiers, 2007, 14(5): 92-103.
[44]
SAROJ, KAVITA. Review: study on simple K-Mean and modified K-Mean clustering technique[J]. International Journal of Computer Science Engineering and Technology, 2016, 6(7): 279-281.
[45]
CUIM Y. Introduction to the K-Means clustering algorithm based on the Elbow method[J]. Accounting, Auditing and Finance, 2020, 1(1): 5-8.
GAOM, WANGG W. A framework for identifying metallogenic intrusions from 2D to 3D based on geological-geophysical datasets in the Wulonggou gold district, Qinghai Province, China[C]// International workshop on Gravity, Electrical & Magnetic Methods and Their Applications. Shenzhen: The Society of Exploration Geophysicists and the Chinese Geophysical Society, 2024: 36-40
[48]
XIONGY H, ZUOR G. A positive and unlabeled learning algorithm for mineral prospectivity mapping[J]. Computers & Geosciences, 2021, 147: 104667.
[49]
XIANGJ, XIAOK Y, CARRANZAE J M, et al. 3D mineral prospectivity mapping with random forests: a case study of Tongling, Anhui, China[J]. Natural Resources Research, 2020, 29(1): 395-414.
[50]
HOT K. Random decision forests[C]// Proceedings of the 3rd international conference on Document Analysis and Recognition. Montreal, Quebec, Canada: IEEE Xplore, 1995, 2: 14-16
GEY, ZHANGZ, CHENGQ, et al. Geological mapping of basalt using stream sediment geochemical data: case study of covered areas in Jining, Inner Mongolia, China[J]. Journal of Geochemical exploration, 2022, 232: 106888.
[56]
GRANEKJ. Application of machine learning algorithms to mineral prospectivity mapping[D]. Vancouver, British; Columbia, Canada: University of British Columbia, 2016.
PLATTJ C. Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods[J]. Advances in Large Margin Classifiers, 1999, 10(3): 61-74.
[59]
CHENT, GUESTRINC. Xgboost: a scalable tree boosting system[C]// 22nd ACM SIGKDD international conference on Knowledge Discovery and Data Mining. New York: Association for Computing Machinery, 2016: 785-794.
XIAY F, LIUC Z, LIY Y, et al. A boosted decision tree approach using Bayesian hyper-parameter optimization for credit scoring[J]. Expert Systems With Applications, 2017, 78: 225-241.
[65]
BERGSTRAJ, BARDENETR, BENGIOY, et al. Algorithms for hyper-parameter optimization[J]. Advances in Neural Information Processing Systems, 2011, 2011: 2546-2554.
[66]
ZHANGZ Q, WANGG W, CARRANZAE J M, et al. Three-dimensional pseudo-lithologic modeling via adaptive feature weighted K-Means algorithm from multi-source geophysical datasets, Qingchengzi Pb-Zn-Ag-Au district, China[J]. Natural Resources Research, 2022, 31: 2163-2179.
[67]
BUCKLANDM, GEYF. The relationship between recall and precision[J]. Journal of the American Society for Information Science, 1994, 45(1): 12-19.
FAWCETTT. An introduction to ROC analysis[J]. Pattern Recognition Letters, 2006, 27: 861-874.
[70]
YANGF F, WANGZ Y, ZUOR G, et al. Quantification of uncertainty associated with evidence layers in mineral prospectivity mapping using direct sampling and convolutional neural network[J]. Natural Resources Research, 2023, 32(1): 79-98.
[71]
CHENG X, HUANGN, WUG P, et al. Mineral prospectivity mapping based on wavelet neural network and Monte Carlo simulations in the Nanling W-Sn metallogenic province[J]. Ore Geology Reviews, 2022, 143: 104765.
[72]
YOUSEFIM, CARRANZAE J M. Data-driven index overlay and Boolean logic mineral prospectivity modeling in greenfields exploration[J]. Natural Resources Research, 2016, 25(1): 3-18.
[73]
YOUSEFIM, CARRANZAE J M. Prediction-area (P-A) plot and C-A fractal analysis to classify and evaluate evidential maps for mineral prospectivity modeling[J]. Computers & Geosciences, 2015, 79: 69-81.
[74]
WANGZ Y, YINZ, CAERSJ, et al. A Monte Carlo-based framework for risk-return analysis in mineral prospectivity mapping[J]. Geoscience Frontiers, 2020, 11: 2297-2308.
[75]
SINGERD A, KOUDAR. Examining risk in mineral exploration[J]. Natural Resources Research, 1999, 8(2): 111-122.
[76]
ZUOR G, WANGZ Y. Effects of random negative training samples on mineral prospectivity mapping[J]. Natural Resources Research, 2020, 29(6): 3443-3455.
[77]
NEALR. Handbook of Markov Chain Monte Carlo, chapter 5: MCMC using Hamiltonian dynamics[M]. New York: CRC Press, 2011.
[78]
HOFFMANM D, GELMANA. The No-U-Turn sampler: adaptively setting path lengths in Hamiltonian Monte Carlo[J]. Journal of Machine Learning Research, 2014, 15(1): 1593-1623.
[79]
GEMANS, GEMAND. Stochastic relaxation, Gibbs distributions and the Bayesian restoration of images[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1984, 6: 721-741.
LIY G, OLDENBURGD W. 3-D inversion of magnetic data[J]. Geophysics, 1996, 61(2): 394-408.
[82]
LIY G, OLDENBURGD W. 3-D inversion of gravity data[J]. Geophysics, 1998, 63(1): 109-119.
[83]
ZHANGZ Q, ZHANGJ J, WANGG W, et al. From 2D to 3D modeling of mineral prospectivity using multi-source geoscience datasets, Wulong gold district, China[J]. Natural Resources Research, 2020, 29: 345-364.