Course material covering applied machine learning techniques for mineral exploration, using real geochemical, geophysical, and remote sensing datasets.
ML_for_geologists.ipynb — The main notebook, structured in sections:
| Section | Topic |
|---|---|
| 1 | Setup — imports, filepaths, data loading |
| 2 | Exploratory data analysis |
| 3 | Unsupervised learning — PCA and K-means clustering |
| 4 | Supervised learning — Random Forest classification |
| 5 | Model evaluation and feature importance |
| 6 | Spatial prediction mapping |
data/
├── vector/
│ ├── geochem.geojson # Geochemical sample points (1,243 samples, 57 elements)
│ ├── lithology.geojson # Lithological polygon map
│ └── mineral occurrences.geojson # Known mineral deposit locations (52 occurrences)
├── raster/
│ ├── geophys/ # Airborne magnetic (AMF) derivatives (RTP, 1VD, 2VD, tilt, etc.)
│ └── spectral/ # Multispectral alteration indices (clay, iron oxide, silica, etc.)
└── assets/
└── image.png
- PCA — dimensionality reduction on multi-element geochemical data
- K-means clustering — unsupervised geochemical domain classification
- Random Forest — supervised binary classification for mineral prospectivity
- Spatial cross-validation — checkerboard train/test split to prevent spatial data leakage
- Feature importance — identifying which geochemical/geophysical features drive predictions
helpers.py contains all supporting functions (data loading, preprocessing, spatial splitting, plotting) kept separate from the notebook to reduce clutter.