The global mineral exploration industry has long faced a fundamental challenge: how to lock in high-potential deposits across vast, unexplored regions with the lowest cost and highest efficiency. Traditional experience-driven and geological analogy methods are not only time-consuming and labor-intensive but also highly subjective and difficult in data fusion, often causing exploration companies to return empty-handed after investing tens or even hundreds of millions of dollars in drilling. Now, an Iranian research team has provided a revolutionary answer using artificial intelligence—covering only 13% of the study area, they successfully locked in almost all high-potential exploration targets in the Kerman porphyry copper belt, paving a new path for the global mineral exploration industry towards data-driven intelligent decision-making.

From "Experience Prospecting" to "Intelligent Target Locking": The Technological Revolution of AI-MPM

On April 30, 2026, a research team from Amirkabir University of Technology (Tehran Polytechnic) and the University of Tehran published a breakthrough result in the prestigious international journal Journal of Earth Science (JCR Q1, published by Springer Nature). For the first time, they systematically proposed a complete AI Mineral Prospectivity Mapping (AI-MPM) strategy. Using the Kerman porphyry copper belt in Iran as a case study, they demonstrated the disruptive capability of artificial intelligence in interpreting complex geological data and delineating low-risk targets.

A "Combination Punch" of Triple Optimization Algorithms Solves the SVM Hyperparameter Problem

Support Vector Machine (SVM) is one of the most mature supervised learning algorithms, but its performance highly depends on the selection of hyperparameters (such as the penalty parameter C and the kernel function parameter γ). Finding the optimal hyperparameters in high-dimensional geological data scenarios typical of mineral exploration is like "looking for a needle in a haystack." The research team, for the first time, synergistically applied three intelligent optimization algorithms—Particle Swarm Optimization (PSO), Grid Search, and Genetic Algorithm (GA)—to the automated hyperparameter tuning of the SVM model. Through rigorous evaluation using multi-layer cross-validation and classification accuracy metrics, this "triple hybrid optimization" strategy significantly improved the model's predictive accuracy and generalization ability, customizing the optimal algorithm parameters for porphyry copper target delineation.

Highlight Two: Pixel-Level Uncertainty Quantification, Moving from "High Probability Correctness" to "Low-Risk Locking"

The biggest flaw of traditional AI mineral prediction is that the model identifies high-potential areas but cannot tell exploration companies the confidence level of this judgment. This research breakthrough developed a pixel-level uncertainty quantification method, calculating the confidence level of the prediction result for each spatial unit. This allows exploration decision-makers to accurately distinguish between "high-potential but high-risk" areas and "high-potential and high-confidence" low-risk targets. This innovation is not just a technical upgrade but a capability leap from "probability prediction" to "risk quantification"—upgrading AI from an "auxiliary inference tool" to an "exploration decision dashboard."

Highlight Three: 13% Area Locks All Potential, Achieving an Order-of-Magnitude Increase in Exploration Efficiency

The Kerman porphyry copper belt is one of the most famous copper metallogenic belts in Iran and globally, with an exploration scope covering thousands of square kilometers. The research team integrated multi-source geospatial data, including geological, geochemical, and geophysical data, to train the AI-MPM model. They ultimately achieved an efficiency figure that shocked the industry: high-quality, low-risk exploration targets accounted for only 13% of the total study area, yet covered almost all high-potential deposit areas within the belt.

This means that across a metallogenic belt spanning thousands of square kilometers, the AI model has compressed the area requiring detailed field investigation for exploration engineers by more than five-sixths. For a mineral exploration industry with drilling budgets often in the tens of millions of dollars, this signifies an exponential decrease in exploration costs and a breakthrough improvement in success rates.

From "Exploration Cost Collapse" to "Global Mining Intelligence"

1. Decision-Making Model Reshaped: From "Geological Intuition" to "Data-Driven"

Traditional mineral exploration relies on the personal cognition and accumulated experience of senior geological experts. This technology, for the first time, systematically applies a triple hybrid optimized integrated AI-MPM strategy to porphyry deposit exploration, marking a shift from "expert-driven" to "algorithm-driven" mineral exploration. It can significantly reduce human bias and the uncertainty of geological interpretation, providing exploration decisions with quantifiable and reproducible scientific evidence.

2. Significant Reduction in Exploration Costs: Precise Navigation Locking 1/6 of the Area

The most direct commercial value of this technology lies in the substantial compression of exploration costs. The main cost expenditure in mineral exploration is field surveys and drilling verification. AI-MPM reduces the exploration area requiring focus to less than one-sixth of traditional methods, meaning the size of field survey teams, drilling workload, and geological sample analysis costs can all decrease correspondingly. Against the backdrop of the global mining industry facing the increasing depletion of high-quality outcrop mines and exponentially rising costs for deep concealed ore exploration, this intelligent target delineation capability will become a core competitiveness for mining companies.

3. Global Transferability and Promotion Value of the New Technological Paradigm

The AI-MPM strategy validated in this study has strong transferability. It is not only applicable to the Kerman copper belt in Iran but can also be promoted to any metallogenic prospect area globally with available multi-source geoscience data. The research team pointed out that future work will further integrate more data sources and deploy cutting-edge machine learning techniques to continuously improve prediction accuracy. As the supply security of global critical minerals (copper, lithium, cobalt, rare earths) increasingly becomes a strategic concern for nations, AI-driven exploration technology is expected to reshape the cost structure of the entire mineral discovery and development process.

Installing an "Intelligent Navigator" for Deep Exploration in Global Mining

The profound value of this research lies in pointing out a viable path for the global mineral exploration industry to move from "labor-intensive" to "intelligence-intensive." As surface outcrop mines become increasingly exhausted and deep concealed deposits and covered area prospecting become the main battlegrounds, traditional geological mapping, soil geochemistry, and geophysical methods are already overwhelmed when facing massive multi-source data. The AI-MPM strategy, empowered by a triple-optimized SVM model combined with pixel-level uncertainty quantification, has for the first time achieved a full-process intelligent closed loop of "data fusion—model decision—risk quantification—target locking" across a metallogenic belt of thousands of square kilometers.

For mining enterprises, this means having a low-cost, high-precision "virtual prospector" before the drill bit finally penetrates deep underground; for the entire mining industry, it heralds the arrival of a paradigm shift from experience-driven to algorithm-driven. As the paper points out, this pioneering research paves the way for the global mining sector to utilize the most advanced artificial intelligence algorithms to explore new mineral resources.