知识-数据联合驱动的可解释智能矿产预测研究:以四川可尔因矿集区为例

doi:10.13745/j.esf.sf.2025.4.63

地学前缘 ›› 2025, Vol. 32 ›› Issue (4): 60-77.DOI: 10.13745/j.esf.sf.2025.4.63

知识-数据联合驱动的可解释智能矿产预测研究:以四川可尔因矿集区为例

李楠¹^,²^,³(), 尹世滔¹^,⁴, 柳炳利², 肖克炎¹, 王成辉¹, 代鸿章¹, 宋相龙¹

1.中国地质科学院矿产资源研究所, 北京 100037
2.成都理工大学数学科学学院, 四川成都 610051
3.中国地质科学院矿产资源研究所, 深地探测与矿产勘查全国重点实验室, 北京 100037
4.中国地质大学(北京), 北京 100083

收稿日期:2025-03-24 修回日期:2025-04-09 出版日期:2025-07-25 发布日期:2025-08-04
作者简介:李楠(1980—),男,研究员,主要从事三维地质建模和矿产资源定量预测评价方面的教学与研究工作。E-mail: linan@cags.ac.cn
基金资助:
国家重点研发计划项目(2021YFC2901905-3);国家重点研发计划项目(2023YFC2906403);中国地质调查局地质调查项目(DD20243233);国家自然科学基金项目(42272347)

A knowledge-data driven interpretable intelligent mineral prediction: A case study of the Keeryin Mineral Concentration Area, Sichuan Province

LI Nan¹^,²^,³(), YIN Shitao¹^,⁴, LIU Bingli², XIAO Keyan¹, WANG Chenghui¹, DAI Hongzhang¹, SONG Xianglong¹

1. Institute of Mineral Resources, Chinese Academy of Geological Sciences, Beijing 100037, China
2. School of Mathematical Sciences, Chengdu University of Technology, Chengdu 610051, China
3. National Key Laboratory of Deep Earth Exploration and Mineral Exploration, Institute of Mineral Resources, Chinese Academy of Geological Sciences, Beijing 100037, China
4. China University of Geosciences (Beijing), Beijing 100083, China

Received:2025-03-24 Revised:2025-04-09 Online:2025-07-25 Published:2025-08-04

摘要/Abstract

摘要：

随着人工智能和大数据技术的迅速发展,基于机器学习的矿产资源智能预测已成为当前研究热点。然而,部分机器学习模型嵌套的复杂非线性网络结构和抽象表达,具有高度不透明的黑盒属性,导致智能预测结果与成矿作用之间缺乏相关解释,降低了预测模型的泛化能力和预测结果的可靠程度。为解决以上问题,本研究提出了知识-数据联合驱动的可解释矿产资源智能预测方法。首先,采用最佳-最差法(BWM)建立了融合先验地质特征权重的集成学习智能预测模型,以强化模型预测效果。之后,使用从全局到局部,从特征到样本的多尺度多维度可解释性方法,解构预测结果,定量评价预测指标重要程度。最后,结合野外验证后的专家指导校正,实现地质找矿知识更新迭代,形成矿床知识嵌入和矿床知识发现完整闭环,进而提升矿产资源智能预测决策过程的透明性和预测结果的可靠性。以四川可尔因矿集区为例进行实验,圈定A类高潜力靶区8处,占总面积的6.58%,其中84%的矿床样本位于高潜力靶区,表明预测方法的稳定性。钠长石频谱、Na₂O+K₂O、环形构造、Li/La和二云母花岗岩依次成为关键预测特征,呈现出明显的有序性,经野外验证,证实其与可尔因伟晶岩型锂矿找矿模型密切相关。

关键词: 矿产资源定量预测, 可尔因矿集区, 知识嵌入, 集成学习, 可解释性, 野外验证

Abstract:

With the rapid advancement of AI and big data, machine learning-based mineral prospectivity mapping has become a research hotspot. However, models with deeply nested, nonlinear structures and abstract representations often exhibit opaque “black-box” characteristics. This lack of interpretability between predictions and metallogenic processes limits model generalization and reliability. To address this, we propose a knowledge-data driven interpretable prediction method. First, the Best-Worst Method(BWM) is used to derive geological feature weights for constructing an ensemble model, enhancing its performance. A multi-scale, multi-dimensional interpretability framework spanning from global to local levels and from feature-level to sample-level interpretations is then applied to deconstruct results and evaluate feature importance. Expert-guided corrections, informed by field validation, further refine the predictions, thereby forming a closed loop of knowledge embedding and discovery. This process improves workflow transparency and result reliability. Applied to the Keeryin area in Sichuan, the method identified eight high-potential Class A targets, occupying only 6.58% of the study area yet containing 84% of the known deposits. Key predictive features included remote sensing spectral response of albite/albite spectral signature, total alkali (Na₂O+K₂O) content, ring structures, Li/La ratio, and two-mica granite—as validated by their strong spatial correlation with known pegmatite-type lithium deposits in the field.

Key words: mineral resource quantitative prediction, Keeryin Mining Area, knowledge embedding, integrated learning, interpretability, field validation

中图分类号:

李楠, 尹世滔, 柳炳利, 肖克炎, 王成辉, 代鸿章, 宋相龙. 知识-数据联合驱动的可解释智能矿产预测研究:以四川可尔因矿集区为例[J]. 地学前缘, 2025, 32(4): 60-77.

LI Nan, YIN Shitao, LIU Bingli, XIAO Keyan, WANG Chenghui, DAI Hongzhang, SONG Xianglong. A knowledge-data driven interpretable intelligent mineral prediction: A case study of the Keeryin Mineral Concentration Area, Sichuan Province[J]. Earth Science Frontiers, 2025, 32(4): 60-77.

图/表 20

参考文献 40

[1]	赵鹏大. 成矿定量预测与深部找矿[J]. 地学前缘, 2007, 14(5): 1-10.
[2]	AGTERBERG F P, BONHAM-CARTER G F. Deriving weights of evidence from geoscience contour maps for the prediction of discrete events[C]// Proceedings of the 22nd APCOM symposium, Berlin, Germany. 1990, 2: 381-395.
[3]	CHENG Q M, AGTERBERG F P. Fuzzy weights of evidence method and its application in mineral potential mapping[J]. Natural Resources Research, 1999, 8(1): 27-35.
[4]	YOUSEFI M, NYKÄNEN V. Introduction to the special issue: GIS-based mineral potential targeting[J]. Journal of African Earth Sciences, 2017, 128: 1-4.
[5]	PORWAL A, CARRANZA E J M. Introduction to the Special Issue: GIS-based mineral potential modelling and geological data analyses for mineral exploration[J]. Ore Geology Reviews, 2015, 71: 477-483.
[6]	左仁广. 基于数据科学的矿产资源定量预测的理论与方法探索[J]. 地学前缘, 2021, 28(3): 49-55. DOI
[7]	陈建平, 李靖, 谢帅, 等. 中国地质大数据研究现状[J]. 地质学刊, 2017, 41(3): 353-366.
[8]	周永章, 陈烁, 张旗, 等. 大数据与数学地球科学研究进展:大数据与数学地球科学专题代序[J]. 岩石学报, 2018, 34(2): 255-263.
[9]	翟明国, 杨树锋, 陈宁华, 等. 大数据时代:地质学的挑战与机遇[J]. 中国科学院院刊, 2018, 33(8): 825-831.
[10]	AGTERBERG F P, BONHAM-CARTER G F, CHENG Q, et al. Weights of evidence modeling and weighted logistic regression for mineral potential mapping[C]//DAVIS J C, HERZFELD U C. Computers in geology:25 years of progress. Oxford: Oxford University Press, 1993:13-32.
[11]	CARRANZA E J M, LABORTE A G. Data-driven predictive mapping of gold prospectivity, Baguio district, Philippines: application of Random Forests algorithm[J]. Ore Geology Reviews, 2015, 71: 777-787.
[12]	向杰, 陈建平, 肖克炎, 等. 基于机器学习的三维矿产定量预测:以四川拉拉铜矿为例[J]. 地质通报, 2019, 38(12): 2010-2021.
[13]	欧阳渊. 西藏冈底斯成矿带西段斑岩型铜矿成矿规律与成矿预测研究[D]. 成都: 成都理工大学, 2020.
[14]	秦耀祖, 吴伟成, 谢丽凤, 等. 基于机器学习的找矿预测模型在湖南岳溪锑矿田的应用[J]. 东华理工大学学报(自然科学版), 2021, 44(1): 28-40.
[15]	张振杰, 成秋明, 杨玠, 等. 机器学习与成矿预测:以闽西南铁多金属矿预测为例[J]. 地学前缘, 2021, 28(3): 221-235. DOI
[16]	HE Z N, QUAN C, WANG S Y, et al. A comparative study of unsupervised deep learning methods for MRI reconstruction[J]. Investigative Magnetic Resonance Imaging, 2020, 24(4):179-195.
[17]	ZUO R, LUO Z, XIONG Y, et al. A geologically constrained variational autoencoder for mineral prospectivity mapping[J]. Natural Resources Research, 2022, 31(3): 1121-1133.
[18]	PARSA M, CARRANZA E J M, AHMADI B. Deep GMDH neural networks for predictive mapping of mineral prospectivity in terrains hosting few but large mineral deposits[J]. Natural Resources Research, 2022, 31(1): 37-50.
[19]	XIONG Y, ZUO R, CARRANZA E J M. Mapping mineral prospectivity through big data analytics and a deep learning algorithm[J]. Ore Geology Reviews, 2018, 102: 811-817.
[20]	ZHANG S, CARRANZA E J M, WEI H, et al. Data-driven mineral prospectivity mapping by joint application of unsupervised convolutional auto-encoder network and supervised convolutional neural network[J]. Natural Resources Research, 2021, 30(2): 1011-1031.
[21]	郑忠超, 王旭东, 王健, 等. 基于多源地学数据的深部找矿预测应用探微[J]. 中国矿业, 2024, 33(增刊1): 605-609.
[22]	杨娜. 多尺度特征融合算法及其在智能成矿预测中的应用[J/OL]. 矿物岩石地球化学通报, 1-14[2025-04-21]. http://kns.cnki.net/kcms/detail/52.1102.P.20250319.1435.001.html.
[23]	WANG Z, ZUO R. Mineral prospectivity mapping using a joint singularity-based weighting method and long short-term memory network[J]. Computers & Geosciences, 2022, 158: 104974.
[24]	ZUO R, XU Y. Graph deep learning model for mapping mineral prospectivity[J]. Mathematical Geosciences, 2023, 55(1): 1-21.
[25]	IBRAHIM M, LOUIE M, MODARRES C, et al. Global explanations of neural networks: mapping the landscape of predictions[C]// Proceedings of the 2019 AAAI/ACM conference on AI, ethics, and society. New York: Association for Computing Machinery (ACM), 2019: 279-287.
[26]	LIU Y, ZHOU K, XIA Q. A MaxEnt model for mineral prospectivity mapping[J]. Natural Resources Research, 2018, 27(3): 299-313.
[27]	PRADHANB, JENA R, TALUKDAR D, et al. A new method to evaluate gold mineralisation-potential mapping using deep learning and an explainable artificial intelligence (XAI) model[J]. Remote Sensing, 2022, 14(18): 4486.
[28]	MOUN, CARRANZA E J M, WANG G, et al. A framework for data-driven mineral prospectivity mapping with interpretable machine learning and modulated predictive modeling[J]. Natural Resources Research, 2023, 32(6): 2439-2462.
[29]	KONG W, CHEN J, ZHU P. Machine learning-based uranium prospectivity mapping and model explainability research[J]. Minerals, 2024, 14(2): 128.
[30]	BAHRAMIY, HASSANI H, MAGHSOUDI A. Spatial modeling for mineral prospectivity using BWM and COPRAS as a new HMCDM method[J]. Arabian Journal of Geosciences, 2022, 15(5): 394.
[31]	左仁广, 成秋明, 许莹, 等. 可解释性矿产预测人工智能模型[J]. 中国科学: 地球科学, 2024, 54(9): 2917-2928.
[32]	张前龙, 周永章, 郭兰萱, 等. 找矿知识图谱的智能化应用: 以钦杭成矿带斑岩铜矿为例[J]. 地学前缘, 2024, 31(4): 7-15. DOI
[33]	王成彬, 王明果, 王博, 等. 融合知识图谱的矿产资源定量预测[J]. 地学前缘, 2024, 31(4): 26-36. DOI
[34]	董宇浩, 王永志, 田江涛, 等. 基于知识图谱的斑岩型铜矿预测研究进展[J]. 地学前缘, 2025, 32(4): 280-290.
[35]	DONGX B, YU Z W, CAO W M, et al. A survey on ensemble learning[J]. Frontiers of Computer Science, 2020, 14(2): 241-258. DOI
[36]	FRIEDMANJ H. Greedy function approximation: a gradient boosting machine[J]. The Annals of Statistics, 2001, 29(5): 1189-1232.
[37]	RIBEIRO M T, SINGH S, GUESTRIN C. “Why should I trust you?”Explaining the predictions of any classifier[C]// Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining. New York: Association for Computing Machinery (ACM), 2016: 1135-1144.
[38]	李鑫, 代鸿章, 王登红, 等. 四川阿坝可尔因锂矿田成矿规律与成矿预测[J]. 岩石学报, 2024, 40(9): 2819-2840.
[39]	付小方, 黄韬, 郝雪峰, 等. 川西花岗细晶岩-伟晶岩型锂矿含矿性评价与示矿标志[J]. 地学前缘, 2023, 30(5): 227-243. DOI
[40]	YIN B, ZUO R, SUN S. Mineral prospectivity mapping using deep self-attention model[J]. Natural Resources Research, 2023, 32(1): 37-56.

类型	控矿要素	地质特征描述	预测要素	预测变量
可尔因花岗伟晶岩型锂矿	构造	构造含矿特征	有利成矿构造	线形构造
	构造	构造含矿特征	有利成矿构造	环形构造
	岩体	岩体含矿特征	成矿有利岩体:二云母花岗岩	钠长石频谱异常
				二云母花岗岩影响域
				Na₂O/K₂O(碱性花岗岩)
			赋矿岩体:锂辉石钠长石型伟晶岩脉	Na₂O+K₂O
	围岩蚀变	有利围岩蚀变	成矿有利蚀变	铁染异常
	围岩蚀变	有利围岩蚀变	成矿有利蚀变	羟基异常
	地球化学	土壤岩屑地球化学	单元素矿化信息	B, Be, La, Li, Nb, P, Rb, Sr, Th
	地球化学	土壤岩屑地球化学	元素异常组合	Li/La, Be-Li-Rb, Nb-Sr-Th

类型	控矿要素	地质特征描述	预测要素	预测变量
可尔因花岗伟晶岩型锂矿	构造	构造含矿特征	有利成矿构造	线形构造
	构造	构造含矿特征	有利成矿构造	环形构造
	岩体	岩体含矿特征	成矿有利岩体:二云母花岗岩	钠长石频谱异常
				二云母花岗岩影响域
				Na₂O/K₂O(碱性花岗岩)
			赋矿岩体:锂辉石钠长石型伟晶岩脉	Na₂O+K₂O
	围岩蚀变	有利围岩蚀变	成矿有利蚀变	铁染异常
	围岩蚀变	有利围岩蚀变	成矿有利蚀变	羟基异常
	地球化学	土壤岩屑地球化学	单元素矿化信息	B, Be, La, Li, Nb, P, Rb, Sr, Th
	地球化学	土壤岩屑地球化学	元素异常组合	Li/La, Be-Li-Rb, Nb-Sr-Th

序号	模型	F₁	序号	模型	F₁
1	SVM	0.83	14	CalibratedClassifierCV	0.62
2	AdaBoost	0.78	15	LabelSpreading	0.62
3	ExtraTrees	0.77	16	GaussianNB	0.61
4	RandomForest	0.71	17	RidgeClassifierCV	0.61
5	DecisionTree	0.70	18	LinearDiscriminantAnalysis	0.60
6	BernoulliNB	0.69	19	ExtraTree	0.60
7	Bagging	0.69	20	Perceptron	0.60
8	Kneighbors	0.68	21	QuadraticDiscriminantAnalysis	0.58
9	Ridge	0.67	22	PassiveAggressive	0.52
10	LinearSVC	0.65	23	NearestCentroid	0.49
11	LabelPropagation	0.65	24	SGD	0.47
12	LogisticRegression	0.65	25	Dummy	0.29
13	LGBM	0.63

序号	模型	F₁	序号	模型	F₁
1	SVM	0.83	14	CalibratedClassifierCV	0.62
2	AdaBoost	0.78	15	LabelSpreading	0.62
3	ExtraTrees	0.77	16	GaussianNB	0.61
4	RandomForest	0.71	17	RidgeClassifierCV	0.61
5	DecisionTree	0.70	18	LinearDiscriminantAnalysis	0.60
6	BernoulliNB	0.69	19	ExtraTree	0.60
7	Bagging	0.69	20	Perceptron	0.60
8	Kneighbors	0.68	21	QuadraticDiscriminantAnalysis	0.58
9	Ridge	0.67	22	PassiveAggressive	0.52
10	LinearSVC	0.65	23	NearestCentroid	0.49
11	LabelPropagation	0.65	24	SGD	0.47
12	LogisticRegression	0.65	25	Dummy	0.29
13	LGBM	0.63

模型	准确率	召回率	F₁分数
Random Forest	0.73	0.66	0.71
ExtraTree	0.79	0.73	0.76
AdaBoost	0.78	0.79	0.78
SVM	0.85	0.78	0.83
Stacking	0.93	0.86	0.91
KE-Stacking	0.96	0.90	0.94

知识-数据联合驱动的可解释智能矿产预测研究:以四川可尔因矿集区为例

A knowledge-data driven interpretable intelligent mineral prediction: A case study of the Keeryin Mineral Concentration Area, Sichuan Province

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 20

参考文献 40

相关文章 4

编辑推荐

Metrics

本文评价

样品编号	取样点位	样品岩性	Li含量/ (μg·g^-1)	Li₂O 含量/%	Be含量/ (μg·g^-1)	Be₂O 含量/%	Rb含量/ (μg·g^-1)	Rb₂O 含量/%	Cs含量/ (μg·g^-1)	W含量/ (μg·g^-1)	Sn含量/ (μg·g^-1)	Co含量/ (μg·g^-1)
GYQ-1	观音桥	花岗伟晶岩	16 989.41	7.31	43.82	0.01	233.28	0.05	20.83	706.92	75.1	86.61
GYQ-2	观音桥	花岗伟晶岩	147.29	0.06	48.35	0.01	531.44	0.11	42.41	493.78	99.6	50.39
GYQ-3	观音桥	花岗伟晶岩	375.93	0.16	52.09	0.01	3 589.41	0.72	231.59	375.48	40.67	37.19
GYQ-4	观音桥	花岗伟晶岩	4 494.31	1.93	98.63	0.03	1 313.01	0.26	96.63	310.37	143.24	25.73
TYH-1	麦地沟	花岗伟晶岩	18.3	0.01	136.85	0.04	1 193.35	0.24	51.33	280.74	30.61	29.03
TYH-2	麦地沟	花岗伟晶岩	36.32	0.02	86.9	0.02	3 725.27	0.75	100.54	422.76	23.65	36.3
TYH-3	麦地沟	花岗伟晶岩	100.79	0.04	335.3	0.09	262.6	0.05	18.21	513.24	50.35	73.12
LB-1	李家沟	花岗伟晶岩	479.53	0.21	9.76	0	588.36	0.12	17.78	303.03	53.16	28.23
LB-1A	李家沟	花岗伟晶岩	219	0.09	4.57	0	284.18	0.06	9.45	165.18	29.24	15.26
LB-2	李家沟	花岗伟晶岩	218.83	0.09	7.55	0	271.5	0.05	15.98	452.35	26.96	43.95
JD-1	加达	花岗伟晶岩	10 815.89	4.65	190.31	0.05	986.25	0.2	88.31	407.24	77.34	41.04
JD-2	加达	花岗伟晶岩	8 882.1	3.82	181.42	0.05	674.14	0.13	65.86	436.17	73.43	45.29
JD-3	加达	花岗伟晶岩	10 758.13	4.63	244.18	0.07	782.67	0.16	80.99	505.51	77.88	57.95
GR-1	高墟	花岗伟晶岩	6 974.02	3	247.01	0.07	1 359.61	0.27	79.56	681.6	107.06	82.68
GR-2	高墟	花岗伟晶岩	4 940.27	2.12	237.19	0.07	1 368.58	0.27	73.95	683.67	60.12	76.79

[1]	孔春芳, 田倩, 刘健, 蔡国荣, 赵杰, 徐凯. 基于集成学习模型与贝叶斯优化算法的成矿预测[J]. 地学前缘, 2025, 32(4): 122-139.
[2]	王琳, 季晓慧, 杨眉, 何明跃, 张招崇, 曾姗, 王玉柱. 基于数据增强和集成学习的矿物图像识别[J]. 地学前缘, 2024, 31(4): 87-94.
[3]	付宇, 曹文庚, 张春菊, 翟文华, 任宇, 南天, 李泽岩. 基于集成学习优化的河套盆地地下水砷风险评估[J]. 地学前缘, 2024, 31(3): 371-380.
[4]	左仁广. 基于数据科学的矿产资源定量预测的理论与方法探索[J]. 地学前缘, 2021, 28(3): 49-55.