Traditional mining exploration techniques require significant effort, including drilling and sample collection, making the process highly challenging and costly. The application of machine learning (ML) in mineral exploration has revolutionized the field by improving efficiency and accuracy in identifying critical raw materials (CRM). This study presents a novel framework that integrates Light Detection and Ranging (LiDAR) and PRISMA hyperspectral data with ML techniques to enhance mineral exploration. By leveraging an ensemble model combining Random Forest (RF) and Multi-Layer Perceptron (MLP), this approach captures complex spatial and spectral patterns, improving the prediction of cobalt, copper, and nickel concentrations. To address the challenge of limited labeled data, synthetic samples were generated using the Gaussian Copula Synthesizer (GCS), enhancing model generalization. The proposed methodology was validated at the A´ ramo mine in Asturias, Spain, demonstrating that the fusion of multispectral and topographical features significantly improves predictive accuracy. The results show that the scalability and robustness of this framework for identifying CRM in geologically significant yet underexplored regions.
