Mineral Prospectivity Machine Learning

Demonstrating data science techniques for mineral exploration using Canadian gold data

Overview

This project demonstrates modern machine learning approaches to mineral prospectivity mapping, inspired by successful crowdsourcing competitions like:

Goldcorp Challenge (2000): Red Lake, Ontario - Led to discovery of $6B+ in gold
Integra Gold Rush (2015-2016): Val-d'Or, Quebec - 1,000+ participants, ML-driven targeting
OZ Minerals Explorer Challenge (2019): Mount Woods, SA - Multi-data fusion approaches

What You'll Learn

✓ How to access and process Canadian geological survey data
✓ Feature engineering for geospatial mineral exploration
✓ Machine learning for imbalanced geological datasets
✓ 3D visualization of prospectivity models
✓ Integration of multiple data types (geology, geophysics, geochemistry)

Project Structure

mineral-prospectivity-ml/
├── README.md                          # This file
├── requirements.txt                   # Python dependencies
├── environment.yml                    # Conda environment
├── data/
│   ├── raw/                          # Downloaded raw data
│   ├── processed/                    # Cleaned, processed data
│   └── README.md                     # Data sources and licenses
├── notebooks/
│   ├── 01_data_acquisition.ipynb     # Download and explore data
│   ├── 02_data_preprocessing.ipynb   # Clean and standardize
│   ├── 03_exploratory_analysis.ipynb # Statistical analysis
│   ├── 04_feature_engineering.ipynb  # Create predictive features
│   ├── 05_ml_modeling.ipynb          # Train ML models
│   └── 06_prospectivity_mapping.ipynb # Generate target maps
├── src/
│   ├── __init__.py
│   ├── data_loader.py                # Data acquisition utilities
│   ├── preprocessing.py              # Data cleaning functions
│   ├── feature_engineering.py        # Feature generation
│   ├── models.py                     # ML model implementations
│   └── visualization.py              # Plotting utilities
└── tests/
    └── test_data_loader.py           # Unit tests

Quick Start

Option 1: Using Conda (Recommended)

# Clone the repository
git clone https://github.com/yourusername/mineral-prospectivity-ml.git
cd mineral-prospectivity-ml

# Create conda environment
conda env create -f environment.yml
conda activate mineral-prospectivity

# Launch Jupyter
jupyter notebook

Option 2: Using pip + virtualenv

# Clone the repository
git clone https://github.com/yourusername/mineral-prospectivity-ml.git
cd mineral-prospectivity-ml

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Launch Jupyter
jupyter notebook

Data Sources

This project uses publicly available Canadian geological data:

Primary Source: Quebec SIGEOM

Drilling Data: 150+ years of exploration drilling
Geochemistry: Rock and soil samples
Geology: Lithology, structures, mineral occurrences
Geophysics: Magnetic, gravity, radiometric surveys

Access: https://sigeom.mines.gouv.qc.ca/

Study Area: Val-d'Or Gold District

Location: Abitibi Greenstone Belt, Quebec
Historical production: 4.5+ million ounces
Active exploration area
Similar to Integra Gold Rush Challenge site

Additional Sources

Natural Resources Canada: National geoscience data
Ontario Geological Survey: Red Lake area (Goldcorp Challenge site)
Geoscience Australia: ASTER geoscience maps (methodology reference)

Key Technologies

Geospatial Analysis

geopandas: Vector data manipulation
rasterio / rioxarray: Raster data processing
pyproj: Coordinate transformations
shapely: Geometric operations

Geoscience Tools

harmonica: Geophysical data processing
verde: Spatial interpolation and gridding
pyvista: 3D geological visualization

Machine Learning

scikit-learn: Random Forest, SVM, ensemble methods
xgboost / lightgbm: Gradient boosting
imbalanced-learn: Handle class imbalance (critical for mineral deposits!)

Data Analysis

pandas / numpy: Data manipulation
xarray: N-dimensional labeled arrays
matplotlib / seaborn / plotly: Visualization

Notebooks

1. Data Acquisition

Download SIGEOM datasets
Define study areas
Load drilling, geology, geophysics data
Initial data quality assessment

Run time: ~10-15 minutes (includes ~500MB download)

2. Data Preprocessing

Handle missing values
Standardize coordinates
Clean geochemical data
Merge multiple data sources

3. Exploratory Data Analysis

Statistical analysis of gold grades
Spatial clustering of deposits
Correlation analysis
Data visualization

4. Feature Engineering

Distance to geological features (faults, contacts)
Geophysical derivatives (TMI, gravity anomalies)
Geochemical ratios and pathfinders
Spatial density calculations

5. Machine Learning Modeling

Handle imbalanced datasets (SMOTE)
Random Forest classification
Model evaluation and cross-validation
Feature importance analysis
Hyperparameter tuning

6. Prospectivity Mapping

Generate prediction grids
Create 3D visualizations
Identify drill targets
Export results

Techniques from Competition Winners

Data Integration

Winning teams consistently emphasized multi-data fusion:

Combine geology, geophysics, geochemistry
Weight different data types appropriately
Handle data at different scales and resolutions

Feature Engineering

Critical features identified by winners:

# Proximity features
- Distance to known deposits
- Distance to favorable lithologies
- Distance to major structures (faults, shear zones)

# Geophysical derivatives
- Vertical derivatives of TMI
- Upward continuation
- Tilt angle
- Analytic signal

# Geochemical indicators
- Element ratios (Au/Ag, Cu/Au)
- Multi-element anomalies
- Pathfinder elements (As, Sb, Te)

Machine Learning Approaches

Random Forest (most common winner):

Handles mixed data types
Robust to outliers
Provides feature importance
Less prone to overfitting

Key considerations:

Class imbalance: Mineral deposits are rare! Use SMOTE or class weights
Spatial autocorrelation: Use spatial cross-validation
Feature scaling: Not always necessary but can help
Ensemble methods: Combine multiple models

Validation Strategy

# Don't use random splits!
# Use spatial cross-validation to avoid data leakage

from sklearn.model_selection import GroupKFold

# Group by geographic location
gkf = GroupKFold(n_splits=5)
for train_idx, test_idx in gkf.split(X, y, groups=spatial_groups):
    # Train and evaluate
    pass

Example Results

After completing the notebooks, you'll be able to:

Load and integrate geological data from multiple sources
Engineer features that predict gold mineralization
Train ML models with >85% accuracy (competition-level)
Generate prospectivity maps showing high-potential areas
Identify drill targets ranked by probability

Contributing

Contributions are welcome! Areas of interest:

Additional data sources (geophysics, remote sensing)
Advanced ML techniques (deep learning, transfer learning)
3D geological modeling integration
Alternative study areas
Documentation improvements

References

Competitions

Tools and Resources

Richard Scott's Mineral Exploration ML: https://github.com/RichardScottOZ/mineral-exploration-machine-learning
Software Underground: https://softwareunderground.org/
Loop3D (3D Modeling): https://loop3d.org/

Scientific Papers

Rodriguez-Galiano et al. (2015): "Machine learning predictive models for mineral prospectivity"
Carranza & Laborte (2015): "Random forest predictive modeling of mineral prospectivity"
Zuo & Xiong (2018): "Big Data Analytics of Identifying Geochemical Anomalies"

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

Quebec Ministry of Natural Resources: For SIGEOM data access
Natural Resources Canada: For national geoscience data
Competition organizers: Goldcorp, Integra Gold, OZ Minerals for demonstrating the power of open innovation
Richard Scott: For the comprehensive mineral-exploration-machine-learning resource compilation

Contact

Questions? Open an issue or reach out via:

GitHub Issues: Project Issues
Email: your.email@example.com

Disclaimer: This project is for educational and research purposes. Always verify geological interpretations with qualified professionals before making exploration decisions.

📍 You are here - but the journey continues... 🚂✈️🚗

This mini-project POC is part a larger umbrella project: 🌍 Around the World (of Analytical Modelling) in 81 Repos - A journey through numerical research, powered by human creativity and AI collaboration

That repo collection is an experiment in rapid numerical research acceleration using the latest LLMs. It demonstrates how humans and AI can collaborate to explore diverse analytical domains—from quantum computing to economics, from statistical mechanics to financial modelling.

Name		Name	Last commit message	Last commit date
data		data
docs		docs
img		img
notebooks		notebooks
sm		sm
src		src
.gitattributes		.gitattributes
PACKAGE_COMPLETE.md		PACKAGE_COMPLETE.md
README.md		README.md
environment.yml		environment.yml
requirements.txt		requirements.txt

Latest commit

History

Folders and files

Repository files navigation