Enhancing mineral exploration in the Kolwezi region, DRC, through remote sensing and ensemble machine learning algorithms

Growing global demand for copper and cobalt requires efficient and sustainable exploration strategies. This study integrates geological reasoning with machine learning to improve exploration insights in the Kolwezi region. It combines remote sensing based geological mapping, analysis of landforms and subsurface structures and assessment of land use and land cover changes at the Musonoi deposit. Three models, Random Forest, Support Vector Machine and Gradient Boosting, were applied to generate prospectivity maps and evaluate spatial relationships between mineralization and geological features. The models achieved high performance with AUC values between 0.89 and 0.91, sensitivity between 0.79 and 0.85 and specificity between 0.87 and 0.91, effectively distinguishing mineralized zones. Results show strong associations between mineralization, structural lineaments, slope breaks and drainage intersections along fault-controlled corridors. Land cover analysis from 1994 to 2024 reveals mining expansion and vegetation loss. This approach enhances exploration efficiency and supports sustainable resource management.