Implementation of machine learning predictive models for targeting gold prospectivity mapping in part of the Ilesha schist belt, southwestern Nigeria

This research leverages innovative machine learning techniques, including artificial neural networks(ANN), support vector machines(SVM), random forests(RF), and Naïve Bayes(NB), to construct a predictive model for gold deposits in part of the Ilesha schist belt. Thirteen predictor maps were used from geophysical, remote sensing, and geological datasets highlighting essential processes in ore formation, such as source characteristics, transport mechanisms, and chemical deposition. Together with 35 gold deposits and non-deposit locations, these datasets were employed to train and test the machine learning models using 10-fold cross-validation techniques. Performance was evaluated with confusion matrices, statistical metrics, and receiver operating characteristic (ROC) curves. All models achieved high predictive accuracy: RF at 98.94%, ANN at 97.80%, SVM at 93.65%, and NB at 90.48%. Key localization factors include magnetic lineament density, structural complexity, SRTM lineament density, analytical signals, and Bouguer gravity anomalies. Using the trained models, prospectivity maps were created, highlighting areas of very-high, high, moderate, and low potential for further research. The Prediction-Area for all the models exceeded 78%, confirming their effectiveness in identifying significant gold deposits in the area. Euler deconvolution suggests geological influences from sill/dyke structures, with magnetic source depths from <-12.96 m to > - 801.67 m and from < −30.82 m to > -1053.94m for structural indices (SI) of 0.5 and 1, respectively while spectral analysis indicates depths of 100m to 1.1 km for shallow sources. A strong correlation exists between gold prospectivity and magnetic sources below 300 m, especially near the Iwaraja/Ifewara fault and lithological boundaries. This study highlights the effectiveness of machine learning models in gold exploration and suggests avenues for future research in similar geological contexts.