Skip to Content
MineralsMinerals
  • Editorial
  • Open Access

25 February 2026

9 Pages

Big Data and AI in Geoscience: From Data to Discovery—A Thematic Overview

,
and
1
School of Earth Sciences and Engineering, Sun Yat-sen University, Zhuhai 519000, China
2
School of Resources and Geosciences, China University of Mining and Technology, Xuzhou 221116, China
3
School of Artificial Intelligence, China University of Geosciences, Beijing 100083, China
*
Author to whom correspondence should be addressed.

1. Introduction

Big data thinking and artificial intelligence are rapidly reshaping how geoscientists analyze, model, and interpret the Earth [1]. This Special Issue focuses on the practical application of big data mining, machine learning, and artificial intelligence in geoscience. It explores the role of big data paradigms in guiding model development, the integration of domain knowledge into AI systems, and the validation of AI methodologies within geoscientific contexts. Through practical examples and case studies, this Special Issue illustrates how these technologies can be applied to complex geoscientific problems, equipping researchers, practitioners, and students with cutting-edge knowledge and tools to harness big data and machine learning—fostering innovation and deepening understanding across geoscience disciplines.
This publication is initiated in commemoration of the 10th anniversary of the Committee on Big Data and AI for Geoscience of the China Society for Mineralogy, Petrology and Geochemistry. Established in 2016 by Chinese scholars, with Professor Yongzhang Zhou as its founding Chair, the committee pioneered organized, specialized research in big data and artificial intelligence applied to mineralogy, petrology, and geochemistry in China [2]. As one of the earliest formal academic organizations worldwide dedicated to big data and AI in the geosciences, it has played a significant leading role in advancing research in this field.
Featuring 17 papers, this Special Issue highlights transformative advances in areas such as mineral prospectivity prediction and metallogenic belt identification, geological data inversion, modeling and deep learning architectures, and geological databases and big data mining. It also introduces knowledge graphs and large language models as emerging tools that enhance data integration, interpretation, and knowledge discovery.
Although this Special Issue does not comprehensively cover the entire field of big data and artificial intelligence in geoscience, it nonetheless reveals prominent research frontiers and distinctive textures at the intersection of AI and geoscience. It is anticipated that the publication of this Special Issue will serve as a valuable reference and catalyst for advancing research in big data and AI in geoscience. By presenting AI-driven geology as a forward-looking paradigm, the Special Issue demonstrates how artificial intelligence can revolutionize traditional geoscientific practices with improved accuracy and deeper insight.

2. Mineral Prospectivity Prediction and Metallogenic Belt Identification

The application of machine learning and deep learning methods to mineral prospectivity prediction and metallogenic belt identification has emerged as a transformative approach in mineral exploration and regional metallogeny. By leveraging high-dimensional geological, geochemical, and geophysical datasets, these data-driven techniques enable the recognition of complex, nonlinear patterns that traditional statistical and empirical models often fail to capture. In mineral prospectivity mapping, supervised learning algorithms—such as random forests, support vector machines, and convolutional neural networks—have been widely adopted to integrate multi-source evidence layers and generate high-resolution predictive maps with improved accuracy and objectivity. For metallogenic belt identification, unsupervised and semi-supervised learning methods, including clustering algorithms and autoencoders, facilitate the delineation of mineralization-related geological assemblages and tectonic settings from regional-scale datasets. More recently, deep learning architectures such as graph neural networks and transformer-based models have shown promise in capturing spatial dependencies and contextual relationships across large geological domains. These approaches not only enhance the efficiency and reproducibility of mineral potential assessments but also support the discovery of subtle mineralization signatures under cover. Collectively, machine learning and deep learning are advancing a new paradigm in predictive mineral system analysis, offering scalable, data-informed frameworks for sustainable resource exploration.
Deep convolutional neural networks (CNNs) applied to geoscientific mineral resource prediction face three critical challenges: (1) model bias caused by imbalanced distributions of ore deposit samples; (2) insufficient global feature extraction due to overreliance on local spatial correlations; (3) weakened discriminative capacity resulting from feature smoothing in deep architectures. To address these issues, Gao et al. [3] in this Special Issue propose a novel T-GCN model that integrates Transformer with graph convolutional networks (GCNs). The model achieves breakthrough performance through three key innovations. First, it leverages the Transformer’s self-attention mechanism to establish a global receptive field, enabling effective capture of long-range geological dependencies. Second, it combines the strengths of GCNs in topological feature extraction to realize multi-scale feature fusion. Third, a feature enhancement module is introduced to mitigate degradation in deep networks. In a case study of the PangXD ore district, the T-GCN model attained a prediction accuracy of 97.27%, outperforming the best comparative model by 3.76 percentage points, and successfully delineated five prospective mineralization zones. These results demonstrate the model’s superior performance and practical value under complex geological conditions.
In the study by Raju et al. [4], a hybrid framework is proposed combining a Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP), a convolutional neural network (CNN) and a Fuzzy-Kernel Extreme Learning Machine (FKELM) to address the challenges of imbalanced and uncertain datasets in mineral exploration. The approach was applied to the G.R. Halli gold prospect in the Chitradurga Schist Belt, Western Dharwar Craton, India, using nine geochemical pathfinder elements. WGAN-GP generated high-quality negative samples, thereby balancing the dataset and reducing overfitting. Compared with support vector machines, gradient boosting, and a baseline CNN, both FKELM (AUC = 0.976, accuracy = 92%) and WGAN-GP + CNN (AUC = 0.973, accuracy = 91%) demonstrated superior performance and produced geologically coherent prospectivity maps. Promising gold targets were delineated, closely aligning with known mineralized zones and geochemical anomalies. This hybrid framework offers a robust, cost-effective, and scalable MPM (mineral prospectivity mapping) solution for structurally controlled geological tracts and data-scarce terrains, and can be integrated with additional geoscience datasets for other complex mineral systems.
Located in the West Qinling Metallogenic Belt of Gansu Province, China, the Xiahe-Hezuo area is characterized by complex regional geology and abundant mineral resources. Several gold–polymetallic deposits have already been identified in this region, highlighting its strong potential for further mineral exploration within the metallogenic belt. Gong et al. [5] addresses the common challenge of small-sample data limitations in geological prediction, the author introduces a Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) to generate high-fidelity geological feature samples, effectively expanding the training dataset. A convolutional neural network (CNN) was used to train and predict on both pre- and post-augmentation data. The experimental results show that, before augmentation, the CNN model’s receiver operating characteristic (ROC) value was 0.9648. After data augmentation with the WGAN-GP, the CNN model’s ROC value improved to 0.9792. Additionally, the CNN model’s classification performance was significantly enhanced, with the training set accuracy increasing by 5% and the test set accuracy improving by 2%, successfully overcoming the issue of insufficient model generalization caused by small sample sizes. The mineralization prediction results based on data augmentation delineate five prospective mineralization targets, whose spatial distribution exhibits strong correlations with known deposits and fault structural belts, confirming the reliability of the predictions. This study validates the effectiveness of data augmentation techniques in MPM and provides a transferable technical framework for MPM in data-scarce regions.
Morocco is rich in Mississippi Valley-Type (MVT) copper deposits. Currently, geochemical surveying is being conducted in the Goulmima region in pursuit of breakthroughs in mineral exploration. Wu et al. [6] focuses on the delineation of prospecting targets in the Goulmima area based on the ongoing 1:100,000 geochemical survey work in Morocco. Their study applies isometric log-ratio (i/r) transformation as a compositional data preprocessing step, followed by Spectrum–Area (S-A) fractal analysis. A random forest (RF) algorithm is then employed for mineral prospectivity modeling. Subsequently, the Concentration–Area (C-A) fractal model is used to further refine the prediction results and delineate high-probability zones, leading to the identification of two prospecting targets. Their results demonstrate the following findings: (1) i/r transformation mitigates the closure effect inherent in compositional data and improves data symmetry, thereby more effectively revealing the spatial structural characteristics of elements; (2) principal component analysis (PCA) applied to the ilr-transformed data successfully identifies two dominant element assemblages—representing a high-temperature hydrothermal environment (Mo–Sn–Ti–W–U) and a low-temperature mineralization environment (CaO–Pb–Zn)—which are consistent with the regional mining history; (3) the S-A multifractal model effectively distinguishes geochemical anomalies from background distributions in the study area, and when integrated with fault buffer zones, provides a reliable basis for mineral prospectivity prediction; (4) the C-A fractal model further classifies the prediction results into high-, medium-, and low-probability zones, and ultimately delineates two prospecting targets.
As is well known, the accuracy of predictive models is often constrained by data quality, feature selection, and algorithmic limitations. To address these, Zhang et al. [7] in this Special Issue investigates the benefits of data-driven feature optimization techniques in enhancing model accuracy. Using the Lhasa region in Tibet as the study area, they applied ensemble learning methods, such as random forest and gradient-boosting tree techniques, to optimize 43 feature variables encompassing geology, geochemistry, and geophysics. The optimized feature variables are then input into a support vector machine (SVM) model to generate a prospectivity map. The performance characteristics of the SVM, RF_SVM, and GBDT_SVM models are evaluated using ROC curves. The results indicate that the feature-optimized GBDT_SVM model achieves superior classification accuracy and prediction effectiveness, demonstrating that feature optimization is a necessary step for mineral prospectivity mapping, as it can significantly improve the performance of mineral prospectivity prediction.
The continuous growth in global demand for mineral resources and the increasing difficulty of mineral exploration have created bottlenecks for traditional mineral prediction methods in handling complex geological information and large amounts of data. To address the evolution of machine learning in large-scale mineral prospectivity prediction since 2016, Fu et al. [8] screened and analyzed 255 high-quality scientific studies through searching the Web of Science core database. These studies cover key areas such as mineral information extraction, target area selection, mineral regularity modeling, and resource potential evaluation. The applied machine learning technologies include random forests, support vector machines, convolutional neural networks, and recurrent neural networks, and have been widely used in the exploration and prediction of various mineral deposits such as porphyry copper, sandstone uranium, and tin. The results indicate a substantial shift within the discipline towards the utilization of deep learning methodologies and the integration of multi-source geological data. There is a notable rise in the deployment of cutting-edge techniques, including automatic feature extraction, transfer learning, and few-shot learning. This study synthesizes the current state and future trajectory of machine learning in large-scale mineral prediction. It seeks to chart the field’s evolution, highlight key research challenges, and identify emerging breakthroughs.

3. Knowledge Graphs and Large Language Models in Geoscience

Knowledge graphs (KGs) and large language models (LLMs) are converging to form a powerful dual-engine framework for intelligent geoscience research. KGs provide explicit, structured representations of geological entities—such as tectonic units, mineralization processes, and deposit types—along with their causal and spatiotemporal relationships. They enable interpretable reasoning and knowledge reuse across heterogeneous datasets. However, traditional KG construction is labor-intensive and difficult to scale. LLMs, particularly domain-adapted models such as GeoGPT and QHGeoGPT, address this bottleneck by automating ontology definition, entity recognition, and relation extraction from massive geoscience text corpora. When integrated with lightweight retrieval-augmented generation (RAG) frameworks such as LightRAG, this synergy enables high-precision knowledge graph construction and dynamic, context-aware question answering. In mineral exploration, such Graph-RAG systems have demonstrated superior performance in terminology understanding and causal chain modeling (e.g., fault systems → fluid migration → ore formation), outperforming general-purpose LLMs by substantial margins. Despite these advances, challenges remain in establishing unified semantic ontologies, ensuring physical consistency and trustworthiness, and achieving seamless integration with numerical models and multi-modal geoscience data. Future directions point toward deep-coupled knowledge–data–model ecosystems that combine symbolic reasoning, neural networks, and generative AI for robust, interpretable, and scalable geoscientific discovery.
In this Special Issue, Feng et al. [9] propose a “pipeline” approach for constructing Earth Science knowledge graphs, aiming to clarify the entire construction process and reduce barriers between data and technology. The approach divides knowledge graph construction into two main stages, each supported by specifically designed functional modules that detail the workflow and enhance reproducibility. Beyond the methodological framework, the authors automatically construct a knowledge graph of iron ore deposits by integrating geographic and geological data using deep learning techniques. This systematic strategy significantly lowers the entry barrier for building geological knowledge graphs, provides transferable methodological support for specific Earth Science disciplines or research objects, and lays a foundation for large-scale Earth Science knowledge graphs that combine crowdsourcing with expert decision making. It also paves the way for the development of intelligent question-answering and decision-making systems spanning the entire Earth Science domain.
In another paper, Ma et al. [10] presents a domain-adapted RAG pipeline that integrates geological knowledge with large language models (LLMs) to support intelligent question answering in the metallogenic domain. Focusing on the Qin–Hang metallogenic belt in South China, a bilingual question–answer (QA) corpus derived from 615 authoritative geological publications is constructed, covering topics such as regional tectonics, ore-forming processes, structural evolution, and mineral resources. Using the ChatGLM3-6B language model and LangChain framework, the corpus is embedded into a semantic vector database via Sentence-BERT and FAISS, enabling dynamic retrieval and grounded response generation. The RAG-enhanced model significantly outperforms baseline LLMs—including ChatGPT-4, Bing, and Gemini—in a comparative evaluation using BLEU, precision, recall, and F1 metrics, achieving an F1 score of 0.8689. The approach demonstrates high domain adaptability and reproducibility. All datasets and codes are openly released to facilitate application in other metallogenic belts. This paper illustrates the potential of LLM-based knowledge engineering to support digital geoscientific research and smart mining.
The integration of heterogeneous and multilingual geoscience texts into coherent knowledge graphs is hindered by semantic inconsistencies arising from terminological variations, diverse linguistic expressions, and data heterogeneity. These challenges significantly impede the construction of reliable knowledge systems for mineral exploration. To address these issue, Qin et al. [11] propose a semantic-aware fusion framework that enables consistent and sustainable integration of mineral exploration knowledge. Built on a standardized geological knowledge schema defining core entities and their interrelations, the framework incorporates an incremental update paradigm via a schema-guided fusion mechanism that detects and resolves semantic conflicts while preserving provenance for traceable evolution. Evaluated on textual sources, the framework achieves an overall triple extraction F1-score of 0.82. Notably, for the critical task of entity extraction, it attains an F1-score of 0.88, outperforming BERT-BiLSTM and BERT-BiLSTM-CRF baselines by up to 11 points. Precision for key metallogenic elements exceeds 0.90. It identifies 1432 conflicts during fusion and generates a refined knowledge graph of 18,204 high-quality de-duplicated triples, retaining 87.3% of inputs. The resulting graph supports downstream applications, including case analysis, visualization, question answering, and mineral prospectivity prediction. Unlike conventional aggregation approaches, this work treats knowledge fusion as a semantically guided dynamic process, enhancing consistency, transparency, and adaptability. It provides a practical pathway toward intelligent and sustainable geoscience knowledge infrastructures.

4. Geological Data Inversion, Modeling, and Deep Learning Architectures

Geological data inversion and modeling aim to reconstruct subsurface structures, physical properties, and mineralization processes from indirect, sparse, and multi-source observations—a fundamentally ill-posed problem. Traditional inversion methods often rely on linear approximations and manual feature engineering, limiting their ability to capture complex, nonlinear geological patterns. Recent advances in deep learning architectures have transformed this landscape. CNNs excel at extracting spatial features from gridded data such as seismic sections and geochemical maps, enabling high-resolution property prediction. Generative adversarial networks (GANs) and variational autoencoders (VAEs) facilitate realistic geological model generation and uncertainty quantification. More recently, physics-informed neural networks (PINNs) integrate governing physical equations into the learning process, ensuring that inversion results honor both observational data and fundamental geological principles. Transformer-based models and graph neural networks (GNNs) further extend these capabilities to irregularly sampled well-log data and complex fault–fracture networks. When coupled with transfer learning and multi-modal fusion strategies, these architectures enable robust, scalable, and interpretable inversion workflows. This synergy between deep learning and geophysical inversion is driving a paradigm shift toward data-driven, physics-aware geological modeling across exploration and basin analysis.
In this Special Issue, He et al. [12] introduces a deep-learning-based object-detection model for intelligent mineral identification, specifically employing the YOLOv8 algorithm. The model was developed with a focus on seven common minerals: biotite, quartz, chalcocite, silicon malachite, malachite, white mica, and pyrite. During the training phase, the model learned to accurately recognize and classify these minerals by analyzing and annotating a large dataset of mineral images. After 258 rounds of training, a stable model was obtained with high performance on key indicators such as precision, recall, mAP50, and mAP50-95, with values stable at 0.91766, 0.89827, 0.94300, and 0.91696, respectively. In the testing phase, using samples provided by the Geological and Mineral Museum at the School of Earth Sciences and Engineering, Sun Yat-sen University, the model successfully identified all test samples, with 83% of them having a confidence level exceeding 87%. Despite some potential misclassifications, the results contribute valuable insights and practical experience to the development of intelligent mineral-recognition technologies.
Dell’Aversana et al. [13] introduces a novel deep neural network (DNN) architecture inspired by natural mycelial networks. Mycelia, the vegetative part of fungi, form extensive underground networks that, in a very efficient way, connect biological entities, transport nutrients and signals, and dynamically adapt to environmental conditions. Drawing inspiration from these properties, MycelialNet integrates dynamic connectivity, self-optimization, and resilience into its artificial structure. This paper explores how mycelial-inspired neural networks can enhance big data analysis, particularly in mineralogy, petrology, and other Earth disciplines, where exploration and exploitation must be efficiently balanced during the process of data mining. The approach is validated by applying MycelialNet to synthetic data first, and then to a large petrological database of volcanic rock samples, demonstrating its superior feature extraction, clustering, and classification capabilities with respect to other conventional machine learning methods.
Geological modeling at the bench scale is often hindered by limited exploration drilling and geophysical wireline data, resulting in uncertainty that undermines production efficiency. Measure-While-Drilling (MWD) systems acquire real-time drilling parameters to evaluate the performance of blast hole drill rigs in mining operations. Early MWD research primarily focused on penetration rates as a means of lithological classification. In this Special Issue, Goldstein et al. [14] investigates artificial intelligence (AI)-based regression models to predict geophysical signatures such as density, gamma, magnetic susceptibility, resistivity, and hole diameter using MWD data. The machine learning (ML) models evaluated include linear regression (LR), decision trees (DTs), support vector machines (SVMs), random forests (RFs), Gaussian processes (GP), and neural networks (NNs). An analytical method was validated for accuracy, and a three-tier experimental method assessed the importance of MWD features, revealing no performance loss when excluding features with less than 2% importance. RF, DTs, and GPs outperformed other models, achieving R2 values up to 0.98 with a low RMSE, while LR and SVMs showed lower accuracy. The NN’s performance improved with larger datasets. This study concludes that the DT, RF, and GP models excel in predicting geophysical signatures. While ML-based methods effectively model relationships in the data, their predictive performance remains inherently constrained by the underlying geological and physical mechanisms. Model selection depends on computational resources and application needs, offering valuable insights for real-time orebody analysis using AI. These findings could be invaluable to geologists who wish to utilize AI techniques for real-time orebody analysis and prediction.
To overcome the limitations of traditional linear inversion methods—particularly their dependence on initial models and the high computational cost of Jacobian matrix calculations, Wang et al. [15] proposed a novel inversion approach for vertical electrical sounding (VES) data based on a backpropagation (BP) neural network optimized by the Particle Swarm Optimization (PSO) algorithm. Two- and three-layer horizontally layered geoelectric models were first constructed to generate training samples. The PSO-BP neural network was then applied to perform inversion tests. Results demonstrated that both BP and PSO-BP neural networks successfully inverted apparent resistivity data into corresponding geoelectric model parameters, validating the correctness of the PSO-BP approach. The method was subsequently applied to field-measured apparent resistivity data. Comparison between inversion results from the PSO-BP network and those from the conventional BP network showed that the PSO-BP network yielded superior performance. These findings further confirm the reliability, effectiveness, and practical applicability of the proposed inversion method. This study provides a novel perspective and a robust alternative for the inversion of vertical electrical sounding data.
Three-dimensional (3D) geological models are fundamental to geological analysis and mineral resource estimation. Conventional on-site survey methods, such as borehole logging, provide localized engineering geological constraints for 3D modeling, yet accurately predicting strata in areas with sparse borehole data remains a persistent challenge. The study by Hang et al. [16] proposes a Transformer-based 3D geological modeling approach tailored to sparse borehole conditions. First, a K-dimensional tree is employed to identify boreholes adjacent to a target point, and a borehole context sequence is constructed using stratigraphic information from these neighboring boreholes. The multi-head attention mechanism of the Transformer model is then applied to capture the relational features between the target point and its adjacent borehole sequence. Finally, trained Transformer encoders are used to predict the stratigraphic category of the target point, with normalized information entropy serving as an uncertainty quantification metric throughout the modeling process. Experimental results demonstrate that the proposed method achieves an accuracy of 0.86, outperforming recurrent neural networks in both predictive accuracy and uncertainty estimation. Moreover, its root mean square error is lower than that of inverse distance weighting and Kriging interpolation. Compared with conventional approaches, the Transformer-based method more accurately delineates the geometry and spatial distribution of geological bodies and better reveals the sedimentary patterns of the study area.
Abulkhair et al. [17] proposed a rapid multivariate resource updating model that integrates new information via ensemble Kalman filter with multiple data assimilations (EnKF-MDA) and multi-Gaussian transformation. Data assimilation enables the optimal merging of models and observations, for which the ensemble Kalman filter (EnKF) has become a widely adopted method. However, resource models in mining typically involve multiple variables characterized by complex multivariate relationships. EnKF is inherently univariate and, even in univariate contexts, may produce slight discrepancies between its estimates and actual observations. This study introduces a methodology for efficiently updating multivariate resource models by combining EnKF-MDA with rotation-based iterative Gaussianization (RBIG). EnKF-MDA enhances the updating process through repeated assimilation of the same data with inflated measurement errors, while RBIG rapidly transforms the data into multi-Gaussian factors. The proposed approach is validated through a real case study involving nine cross-correlated variables. The integration of EnKF-MDA and RBIG effectively improves the accuracy of resource model updates, reduces uncertainty, and preserves the underlying multivariate relationships.

5. Geological Databases and Big Data Mining

Geological databases serve as centralized repositories for vast and diverse geoscientific data, including lithological descriptions, geochemical assays, geophysical surveys, mineral deposit records, and remote sensing imagery. As the volume, variety, and velocity of such data continue to grow due to advances in sensor technologies, automated drilling, and digital archiving, traditional data processing methods are no longer sufficient to extract meaningful insights. Big data mining has therefore emerged as a critical paradigm in the geosciences, leveraging machine learning, artificial intelligence, and high-performance computing to uncover hidden patterns, correlations, and trends within massive, multi-source datasets. These techniques enable more accurate predictive modeling for mineral exploration, reservoir characterization, and geohazard assessment, while also supporting real-time decision making in mining operations. Moreover, the integration of heterogeneous data types into interoperable geological databases facilitates cross-disciplinary research and enhances the reproducibility of scientific findings. Despite ongoing challenges such as data heterogeneity, inconsistent standards, and computational scalability, the fusion of geological databases with big data mining is driving a transformative shift toward data-driven geoscience. This synergy not only improves resource discovery and extraction efficiency but also contributes to more sustainable environmental management and risk mitigation.
Traditional gemstone databases have three major limitations: an absence of standardized data schemas, incomplete core datasets (e.g., records of synthetic and treated gemstones and inclusion characteristics), and poor data interoperability. These deficiencies hinder the application of advanced technologies, such as machine learning (ML) and AI techniques. Zhang et al. [18] in this Special Issue proposes a framework for a new gemstone database based on standardization (FAIR principles), integration (blockchain technology), and dynamism (real-time updates). This framework could transform the gemstone industry, shifting it from “experience-driven” to “data-driven” practices. Powered by big data technology, this novel database may revolutionize gemological research, jewelry authentication, market transactions, and educational outreach, fostering innovation in academic research and practical applications.
Clustering, which groups large datasets based on distance metrics, is an essential method in multidimensional data analysis. Another paper by Chang et al. [19] introduces the Two-Step Clustering (TSC) approach offers advantages by handling both categorical and continuous variables and automatically determining the optimal number of clusters. The authors applied the TSC method to mineral prediction in the northeastern margin of the Jiaolai Basin in the following ways: (1) converting residual gravity and magnetic anomalies into categorical variables using Ward clustering; (2) transforming 13 stream sediment elements into independent continuous variables through factor analysis. The results showed that clustering is sensitive to categorical variables and performs better with fewer categories. When variables share similar distribution characteristics, consistency between geophysical discretization and geochemical boundaries also influences clustering results. In their study, the (3 × 4) and (4 × 4) combinations yielded optimal clustering results. Cluster 3 was identified as a favorable zone for gold deposits due to its moderate gravity, low magnetism, and the enrichment in F1 (Ni–Cu–Zn), F2 (W–Mo–Bi), and F3 (As–Sb), indicating a multi-stage, shallow, hydrothermal mineralization process. This study demonstrates the effectiveness of combining Ward clustering for variable transformation with TSC for the integrated analysis of categorical and numerical data, confirming its value in multi-source data research and its potential for further application.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Zhou, Y.; Zhang, L.; Zhang, A.; Wang, J. Big Data Mining and Machine Learning in Earth Sciences; Sun Yat-sen University Press: Guangzhou, China, 2018; pp. 1–269. [Google Scholar]
  2. Zhou, Y.; Xiao, F. Overview: A glimpse of the latest advances in artificial intelligence and big data geoscience research. Earth Sci. Front. 2024, 31, 1–6. [Google Scholar]
  3. Gao, L.; Gopalakrishnan, G.; Nasri, A.; Li, Y.; Zhang, Y.; Ou, X.; Xia, K. Transformer–GCN Fusion Framework for Mineral Prospectivity Mapping: A Geospatial Deep Learning Approach. Minerals 2025, 15, 711. [Google Scholar] [CrossRef] [Scilit]
  4. Raju, P.V.S.; Mudili, V.S.; Ganivada, A. A Hybrid Framework for Detecting Gold Mineralization Zones in G.R. Halli, Western Dharwar Craton, Karnataka, India. Minerals 2025, 15, 1125. [Google Scholar] [CrossRef] [Scilit]
  5. Gong, J.; Li, Y.; Xie, M.; Kong, Y.; Tang, R.; Li, C.; Wu, Y.; Wu, Z. Mineral Prospectivity Mapping in Xiahe-Hezuo Area Based on Wasserstein Generative Adversarial Network with Gradient Penalty. Minerals 2025, 15, 184. [Google Scholar] [CrossRef] [Scilit]
  6. Wu, Y.; Sun, L.; Qu, Z.; Yu, W.; Zhang, P.; Jing, G.; Shen, P.; Tian, S.; Wang, Q.; Liu, H.; et al. Utilizing Multifractal and Compositional Data Analysis Combined with Random Forest for Mineral Prediction in Goulmima, Morocco. Minerals 2025, 15, 222. [Google Scholar] [CrossRef] [Scilit]
  7. Zhang, H.; Xie, M.; Dan, S.; Li, M.; Li, Y.; Yang, D.; Wang, Y. Optimization of Feature Selection in Mineral Prospectivity Using Ensemble Learning. Minerals 2024, 14, 970. [Google Scholar] [CrossRef] [Scilit]
  8. Fu, Z.; Zheng, X.; Yan, Y.; Xu, X.; Zhou, F.; Li, X.; Zhou, Q.; Mai, W. The Evolution of Machine Learning in Large-Scale Mineral Prospectivity Prediction: A Decade of Innovation (2016–2025). Minerals 2025, 15, 1042. [Google Scholar] [CrossRef] [Scilit]
  9. Feng, Q.; Zhao, T.; Liu, C. A “Pipeline”-Based Approach for Automated Construction of Geoscience Knowledge Graphs. Minerals 2024, 14, 1296. [Google Scholar] [CrossRef] [Scilit]
  10. Ma, J.; Zhou, Y.; He, L.; Zhang, Q.; Bilal, M.A.; Zhang, Y. Enhancing Geological Knowledge Engineering with Retrieval-Augmented Generation: A Case Study of the Qin–Hang Metallogenic Belt. Minerals 2025, 15, 1023. [Google Scholar] [CrossRef] [Scilit]
  11. Qin, Y.; Yang, H.; Cui, L.; Zhang, Y.; Feng, G.; Qiao, Y.; Yao, Y. Semantic-Aware Fusion of Mineral Exploration Knowledge Streams Towards Dynamic Geological Knowledge Graphs. Minerals 2025, 15, 1257. [Google Scholar] [CrossRef] [Scilit]
  12. He, L.; Zhou, Y.; Zhang, C. Application of Target Detection Based on Deep Learning in Intelligent Mineral Identification. Minerals 2024, 14, 873. [Google Scholar] [CrossRef] [Scilit]
  13. Dell’Aversana, P. A Biological-Inspired Deep Learning Framework for Big Data Mining and Automatic Classification in Geosciences. Minerals 2025, 15, 356. [Google Scholar] [CrossRef] [Scilit]
  14. Goldstein, D.; Aldrich, C.; Shao, Q.; O’Connor, L. Unlocking Subsurface Geology: A Case Study with Measure-While-Drilling Data and Machine Learning. Minerals 2025, 15, 241. [Google Scholar] [CrossRef] [Scilit]
  15. Wang, Y.; Gu, G.; Wu, Y.; Wang, S.; Niu, X.; Xu, Z.; He, H.; Lin, X.; Cao, L. Inversion of Vertical Electrical Sounding Data Based on PSO-BP Neural Network. Minerals 2025, 15, 925. [Google Scholar] [CrossRef] [Scilit]
  16. Hang, Z.; Xue, T.; Chen, J.; Shi, Y.; Yin, Z.; Cui, Z.; Zhou, G. A 3D Geological Modeling Method Using the Transformer Model: A Solution for Sparse Borehole Data. Minerals 2025, 15, 301. [Google Scholar] [CrossRef] [Scilit]
  17. Abulkhair, S.; Dowd, P.; Xu, C.; Stewart, P. Rapid Updating of Multivariate Resource Models Based on New Information Using EnKF-MDA and Multi-Gaussian Transformation. Minerals 2025, 15, 424. [Google Scholar] [CrossRef] [Scilit]
  18. Zhang, Y.; Shi, G. Big Data and AI-Enabled Construction of a Novel Gemstone Database: Challenges, Methodologies, and Future Perspectives. Minerals 2025, 15, 1149. [Google Scholar] [CrossRef] [Scilit]
  19. Chang, X.; Zhang, M.; Chen, L.; Zhang, S.; Ren, W.; Zhang, X. Application of Clustering Methods in Multivariate Data-Based Prospecting Prediction. Minerals 2025, 15, 760. [Google Scholar] [CrossRef] [Scilit]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.