Interpretable machine learning for mineral prospectivity mapping in the Qulong–Jiama district, Tibet, China

Abstract

Machine learning (ML) models have been successfully adopted to delineate prospecting regions for a specific type of mineral deposit in data-driven mineral prospectivity mapping (MPM). The primary objective of the ML-based MPM is to effectively integrate multi-source mineral exploration information and enhance its predictive capability and precision. Prior studies demonstrated that one may achieve an improved performance MPM by using models trained by exploration targeting criteria closely associated with mineral deposits, along with coherent training samples with similar multivariate spatial data signatures. Random Forest (RF), Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP) are employed to conduct mineral prospectivity maps in this study. By employing the Permutation Feature Importance and SHAP (SHapley Additive exPlanations) analysis from a global interpretable perspective, this study successfully identified the main impactful evidence layers contributing to mineralization predictions. Furthermore, the application of local interpretability through SHAP analysis facilitated the identification of regions where evidence layers provided consistent contributions to predictions, demonstrating a similar multivariate spatial data signature. By employing interpretable machine learning techniques, not only is the explainability of the model’s predictions significantly improved, but the performance of MPM is also markedly enhanced. Utilizing models trained on exploration targeting criteria closely associated with mineral deposits, along with coherent training samples characterized by similar multivariate spatial data signatures, the final probability map achieved an AUC value of 0.970 and exhibited strong spatial correlation with known deposits. This approach effectively delineates high-probability areas, thereby optimizing the identification of potential mineralization zones and providing guidance for future copper exploration efforts in the Qulong-Jiama district.