A Machine Learning Framework for the Investigation of Energy-Critical Mineralized Geologic Structures from Gravity and Magnetic Datasets: Implications for Sustainable Exploration

Abstract

Mineral exploration faces challenges from complex geological architectures and subtle geophysical expressions of mineralization. This study proposes a joint gravity–magnetic machine learning framework integrating magnetic anomaly, analytic signal, gravity, and Source Parameter Imaging depth estimates to enhance mineralization prediction in Nigeria's Middle Benue Trough and adjoining basement terrain. Five supervised algorithms were evaluated, with Random Forest achieving the highest performance (accuracy=0.954, precision=0.889, recall=0.819, macro-F1=0.850, ROC-AUC=0.978). Correlation analysis revealed low feature redundancy, while unsupervised clustering confirmed structural partitions consistent with mapped fault systems. Ablation studies identified analytic signal as the most influential predictor; however, gravity and depth features contributed essential complementary information, increasing predictive accuracy by over 10% when combined with magnetic data. The resulting mineralized structure map aligns with conventional interpretations while delineating previously unrecognized targets in subdued magnetic response areas. This framework provides an objective, geologically defensible tool for energy-critical mineral targeting, demonstrating substantial improvements over traditional methods while minimizing environmental disturbance through enhanced exploration efficiency. The approach is directly applicable to sustainable resource development in similar terranes worldwide.