Detecting Low-Risk Exploration Targets from Automated Tuned Machine Learning Algorithms: Hybrid Particle Swarm Optimization, Grid Search and Genetic Optimization Algorithms with Support Vector Machine

This study brings a revolutionary approach for defining low-risk exploration opportunities in mineral exploration. By harnessing the power of artificial intelligence (AI), it introduces an innovative strategy, known as AI mineral prospectivity mapping (AI-MPM), which aims to discover untapped mineral resources in unexplored territories. It introduces an innovative methodology that examines the uncertainty surrounding SVM-based predictive models. The study focuses on exploring "Cu porphyry mineralization in Kerman belt, Iran," an area rich in untapped mineral wealth. The researchers utilize cutting-edge optimization algorithms to optimize the performance of SVM algorithm and deduce optimal hyperparameters (e.g., Grid search, PSO and GA). Diverse datasets are synthesized to gain a comprehensive understanding of the study area. The efficiency of auto tuned SVM-based prospectivity models is evaluated using various metrics, including K-fold cross-validation, confusion matrix, and classification accuracy indices values. The research successfully quantifies uncertainty at an individual cell level, enabling accurate identification of low-risk exploration targets. The approach demonstrates remarkable efficiency by covering only 13% of the entire study area. This research paves the way for integrating additional data sources and deploying advanced machine learning techniques to enhance predictive accuracy. Overall, it highlights the potential of SVM-based models in identifying low-risk exploration targets and advancing mineral exploration.