Abstract
This study presents an unsupervised workflow for delineating prospectivity zones for stratiform copper-cobalt mineralization in the southern margin of the Central African Copperbelt, Zambia, using only airborne geophysical data (gravity, magnetics, radiometry, terrain). Three clustering algorithms-K-Means, Fuzzy C-Means (FCM), and Self-Organizing Maps (SOM)-were applied, followed by consensus clustering to isolate robust target zones. Key geophysical filters (tilt derivative, total horizontal derivative, analytic signal) and radiometric ratios (U/Th, U/K) were computed and compared across clusters. The most prospective cluster was validated against literature-calibrated geophysical thresholds and was found to match known exploration criteria and regional structural trends. A final probabilistic prospectivity map was generated from FCM membership, classifying targets into three confidence levels. While lack of drill-hole data prevents quantitative accuracy assessment, this approach demonstrates that fully unlabeled machine learning on airborne data can effectively guide early-stage exploration in data-sparse regions. The proposed workflow offers a reproducible framework for AI-driven prospectivity mapping in frontier terrains.