Application of multispectral satellite imagery and artificial intelligence-based machine learning for mineral exploration: A case study of Kossanto, Western Mali

The discovery of new gold deposits increasingly requires innovative exploration methods capable of integrating complex geoscientific datasets. This study presents an integrated approach combining multispectral satellite imagery, Geographic Information Systems (GIS), and artificial intelligence (AI)-based machine learning for predictive mineral prospectivity mapping in the Kossanto area, western Mali. The datasets used include Sentinel-2 MSI and Landsat 8 OLI imagery, together with geological, structural, and topographic data. Several remote sensing techniques, including false-color composites, spectral band ratios, Principal Component Analysis (PCA), and hydrothermal alteration indices, were applied to extract spectral signatures associated with lithological units and mineralized zones. The derived variables were subsequently integrated into three machine learning models: Random Forest (RF), Support Vector Machine (SVM), and Convolutional Neural Network (CNN). The results demonstrate the ability of these models to capture complex relationships between spectral, geological, and structural parameters for predicting gold mineralization. Among the evaluated algorithms, the CNN model achieved the highest predictive performance, followed by Random Forest and SVM, confirming the effectiveness of machine learning techniques in mineral prospectivity mapping. The resulting prospectivity map highlights priority exploration targets closely associated with major tectonic structures, lithological contacts, and hydrothermal alteration zones. The findings demonstrate that integrating multispectral remote sensing with artificial intelligence provides a robust and cost-effective framework for optimizing gold exploration, reducing exploration costs, and improving target selection within the West African Craton.