Document Type : Original Research Paper

Authors

• Hamidreza Ramazi ¹
• Moslem Jahantigh ²

¹ Amirkabir University of Technology

² Mine engineering, Mine, Amir Kabir university of technology, Tehran, Iran

Abstract

This study focuses on developing and rigorously evaluating two optimized Support Vector Machine (SVM) models—Genetic Algorithm-SVM (GA-SVM) and Particle Swarm Optimization-SVM (PSO-SVM)—for Porphyry Copper Mineral Potential Mapping (MPM) in the Shahr-e-Babak region. Utilizing Dataset #1, selected based on the highest average Area Under the Curve (AUC) via the ROC method, the models were implemented and their hyperparameters optimized using GA and PSO techniques, respectively. Performance was assessed using standard classification metrics derived from the Confusion Matrix, as well as spatial prediction indicators. The GA-SVM model consistently outperformed the PSO-SVM model across all metrics. Specifically, GA-SVM achieved an overall Accuracy of 89% compared to 82% for PSO-SVM. Furthermore, spatial analysis showed the GA-SVM model possessed a superior predictive efficiency, evidenced by a higher Normalized Density value of 3.6 compared to 3.05 for PSO-SVM, and successfully predicted a greater percentage (72% vs. 58%) of known mineral indices within its prospective area. The findings conclusively establish the GA-SVM methodology as the superior framework for producing an accurate and efficient mineral potential map in the study area.

Keywords

• SVM
• Genetic algorithm
• Particle Swarm Optimization
• porphyry
• Shahr-e-Babak

Main Subjects

• Exploration