A New Method to Evaluate Gold Mineralisation-Potential Mapping Using Deep Learning and an Explainable Artificial Intelligence (XAI) Model

Geoscientists have extensively used machine learning for geological mapping and exploring the mineral prospect of a province. However, interpretation results becomes challenging due to complexity models. This study uses convolutional neural network (CNN) Shapley additive explanation (SHAP) estimate potential locations gold mineralisation in Rengali Province, tectonised mosaic volcano-sedimentary sequences juxtaposed at interface Archaean cratonic segment north Proterozoic granulite provinces Eastern Ghats Belt India. The objective is integrate multi-thematic data involving geological, geophysical, mineralogical geochemical surveys on 1:50 K scale with aim prognosticating mineralisation. available utilised during integration include aero-geophysical (aeromagnetic aerospectrometric), (national mapping), ground geophysical (gravity), satellite gravity, remote sensing (multispectral) National Geomorphology Lineament Project structural lineament maps obtained from Geological Survey India Database. CNN model has an overall accuracy 90%. SHAP values demonstrate that major contributing factors are, sequential order, antimony, clay, lead, arsenic content magnetic anomaly modelling. Geochemical pathfinders, including factors, high importance, followed by shear zones mapping. According results, central parts area, river valley, higher prospects than surrounding areas. Gold possibly associated intermediate metavolcanics along zone, which later intruded quartz veins northern part Province. work intends known occurrences respect multiple themes so can be replicated