Recent advances in artificial intelligence (AI) methods have significantly enhanced mineral prospectivity mapping. AI-based algorithms offer high capability for regional-scale mapping of underexplored zones with high mineral potential. However, there are still significant methodological challenges, particularly in integrating multidimensional, heterogeneous geospatial datasets and handling their inconsistencies with sparse and spatially clustered distributed known mineral deposits. This paper presents a novel framework to address these challenges. We developed a training dataset comprising 69 training features derived from geological, geophysical, and lithospheric raster grids. Vector-based geological features were systematically converted into raster grids, where each pixel encodes the minimum distance to the nearest structural and lithological boundaries. Therefore, one can capture the influence of structural and lithological proximity on metallogenic zones. The training target was generated by combining the spatial distribution of seven major metallic mineral deposits and converting their spatial locations into a continuous raster grid of spatial density of mineral deposits. Seven machine learning algorithms with their 24 subtypes were used to predict the spatial density of mineral deposits. Among them, the ensemble bagged trees method showed optimum prediction performance by achieving the lowest root mean square error and the highest coefficient of determination. The optimized model was applied to calculate a predictive mineral prospectivity map across the Iranian plateau. To pinpoint underexplored high-potential zones, residual spatial density anomalies were calculated by subtracting observed deposit occurrence spatial densities from the predicted prospective grid. The residual spatial density anomalies revealed several promising areas, such as the Malayer–Isfahan Pb–Zn zone along the Zagros suture zone. The residual anomalies showed significant potential extended southward of the KaraDagh Cu zone in NW Iran. The results also indicated high-potential zones in central and eastern Iran, notably near the Bafgh, Nehbandan–Ferdous, and Jiroft–Shahrebabak metallogenic zones. Regional-scale AI-aided regression analysis enhanced our understanding of mineral deposit distribution across the Iranian plateau. This insight provides a strategic foundation for future national-scale exploration programs by improving efficiency, reducing risk and cost, and narrowing the area of detailed exploration. |
