Search

AI-driven mineral prospectivity mapping (MPM) is a valid and increasingly accepted tool for delineating the targets of mineral exploration, but it suffers from noisy and unrepresentative input features. In this study, a set of fractal and multifractal methods, including box-counting calculation, concentration–area fractal modeling, and multifractal analyses, were employed to excavate the underlying nonlinear mineralization-related information from geological features. Based on these methods, multiple feature selection criteria, namely prediction–area plot, K-means clustering, information gain, chi-square, and the Pearson correlation coefficient, were jointly applied to rank the relative importance of ore-related features and their fractal representations, so as to choose the optimal input feature dataset readily used for training predictive AI models. The results indicate that fault density, the multifractal spectrum width (∆α) of the Yanshanian intrusions, information dimension (D1) of magnetic anomalies, correlation dimension (D2) of iron-oxide alteration, and the D2 of argillic alteration serve as the most effective predictor features representative of the corresponding ore-controlling elements. The comparative results of the model assessment suggest that all the AI models trained by the fractal datasets outperform their counterparts trained by raw datasets, demonstrating a significant improvement in the predictive capability of fractal-trained AI models in terms of both classification accuracy and predictive efficiency. A Shapley additive explanation was employed to trace the contributions of these features and to explain the modeling results, which imply that fractal representations provide more discriminative and definitive feature values that enhance the cognitive capability of AI models trained by these data, thereby improving their predictive performance, especially for those indirect predictor features that show subtle correlations with mineralization in the raw dataset. In addition, fractal-trained models can benefit practical mineral exploration by outputting low-risk exploration targets that achieve higher capturing efficiency and by providing new mineralization clues extracted from remote sensing data. This study demonstrates that the fractal representations of geological features filtered by multi-criteria feature selection can provide a feasible and promising means of improving the predictive capability of AI-driven MPM.

The development of mineral prospectivity mapping (MPM), which aims to outline and prioritize mineral exploration targets, has been spurred by advances in data-driven machine learning algorithms. Supervised data-driven MPM is a typical few-shot task, suffering from a scarcity of labeled data, the over-fitting of models and an uncertainty of predictions. The main objective of this contribution is to propose a robust framework of few-shot learning (FSL), combining data augmentation and transfer learning to enable the generation of prospectivity models with excellent predictive efficiency and low uncertainty. The mineral systems approach was used to transfer a conceptual mineral system into mappable exploration criteria. Synthetic minority over-sampling technique (SMOTE) was employed to augment and balance the labeled dataset, allowing for model pre-training with the large synthetic training dataset of a source domain. The knowledge derived from pre-trained models was then transferred to the target domain by fine-tuning, and the prospectivity model was generated in light of over-fitting and uncertainty assessments. The proposed FSL framework was applied to tungsten prospectivity mapping in southern Jiangxi Province. The results indicated that the SMOTE-ed balanced dataset boosted the classification accuracy in the training process. The FSL models yielded an arch-shaped prediction point pattern which was favorable for focusing potential targets with high probability and low uncertainty. The FSL models achieved a high predictive performance (test AUC = 0.9172) and the lowest quantitative over-fitting value compared to the models derived from the benchmark algorithms of random forest and support vector machine. Four levels of potential targeting zones, considering both predictive efficiency and uncertainty, were extracted from the resulting FSL prospectivity map. The final high-potential and low-risk exploration targets only cover 4.27% of the area, but capture 41.53% of known tungsten deposits, thus achieving a superior predictive performance. This study highlights the capability of FSL framework to control over-fitting and generate high-confidence exploration targets with low levels of uncertainty.