Suitability of available regional geoscientific data for mapping areas with concealed orogenic gold mineralization using machine and deep learning algorithms in the Amalia–Kraaipan greenstone Belt, South Africa

The Amalia–Kraaipan Greenstone Belt (AKGB) is one of the exploration-immature greenstone belts in South Africa, with one currently operational mine, the Kalgold Mine, owned and managed by Harmony Ltd. The AKGB is semi-concealed by younger sediments, with sporadic and limited outcrops as inliers, making the terrane unattractive for mineral exploration, especially by smaller emerging companies, due to potentially higher exploration and mining costs. In the current study, we have adopted five data-driven integration algorithms to assess the suitability of regional legacy geoscientific datasets and the efficacy of these algorithms for mineral prospectivity mapping of concealed orogenic gold mineralization. The algorithms include unsupervised principal component analysis (PCA), a deep autoencoder (DAE), and supervised random forest (RF), support vector machines (SVM), and convolutional neural networks (CNN). All the models were able to delineate known mineral deposits and occurrences. They also demonstrated high predictive ability based on receiver operating characteristic (ROC), success–rate plot, and prediction–area (P–A) plot. The models’ validation tools showed complementary results, and all demonstrated that CNN outperformed other models with an area under the curve (AUC) of 0.835 and a higher P–A intersection point, followed by the SVM (0.795), PCA (0.764), RF (0.755), and DAE (0.545), respectively. The results supported that both the regional geoscientific datasets and data integration techniques are usable in mapping orogenic gold mineralization within the Amalia–Kraaipan Greenstone Belt. The outlined potential mineral targets are recommended for further investigation and exploration of mineral deposits. We also recommend the acquisition of high-resolution datasets for mapping subsurface geology and detailed structural forms.