The increasing global demand for strategic materials underscores the need for advanced methods to balance mineral resource extraction with environmental risk assessments. In this context, this research aims to develop a Bayesian methodology for the probabilistic prediction of mineral-rich deposits while integrating a qualitative assessment of the environmental risks associated with mineral extraction. For this purpose, the selected area to validate the proposed Automated Machine Learning (AutoML) tool is the former Sn–W Mata da Rainha mine, part of the Góis–Panasqueira–Segura Sn–W metallogenic belt in Central Portugal. After mining activities ceased, 706 stream sediment samples were collected, enabling the identification of both remaining mineral-enriched zones and potentially contaminated areas. As mentioned, the geochemical results were analyzed through AutoML and Bayesian modeling. First, an unsupervised Bayesian machine learning approach explored unconstrained relationships between variables. Then, a supervised method generated a robust model to predict Sn and W—key elements in the mining context. As a result, Fe, Cr, P, and V emerged as key attributes due to their role position in the Bayesian model’s exploratory structure and high predictive value. In conclusion, the proposed approach enables data-driven decision-making in economically viable mining areas while mitigating associated environmental risks.