Berkay Ersus, MSc
B.Ersus1, C. Farquharson1
1Department of Earth Sciences, Memorial University, St. John's, Newfoundland, Canada
Mineral prospectivity mapping (MPM) is an essential tool for identifying high-potential zones for mineralization, providing a data-driven approach to enhance exploration targeting. This study focuses on the Baie Verte Peninsula, Newfoundland, with its complex geological history and significant gold occurrences. The region hosts two primary gold deposits: orogenic vein-hosted gold and disseminated gold in ore zones, both associated with structural features and lithological controls. This research integrates diverse data into a unified analysis framework by leveraging open-source datasets from the Federal Government of Canada.
Using a support vector machine (SVM) algorithm, predictive modelling was performed on key features, including fault proximity, geochemical anomalies, geological unit interactions, etc., critical indicators of gold mineralization. The results yielded a prospectivity map for potential gold deposits and provided a reliable tool for exploration decision-making.
As part of this project, a new software solution tailored for mineral prospectivity mapping was developed, which supports two machine learning algorithms (SVM and Random Forest) and one deep learning algorithm (Convolutional Neural Networks, CNN). While this software offers flexibility in algorithm selection, the current study exclusively utilizes the SVM algorithm for its robustness and suitability to the dataset. This research demonstrates the effectiveness of combining AI-driven methods with public geoscientific data to improve the efficiency and precision of mineral exploration. This study contributes to the broader adoption of sustainable, data-driven practices in the mining industry and aligns with the goals of enhancing resource exploration in Canada.