This model predicts where gold occurs in Georgia from public geologic data. A prediction is not a result. The validation is the result, so it is on this page too.
Spatial-block cross-validation, five folds, XGBoost. Two rows are emphasized. They answer different questions, and quoting one alone would be dishonest.
| Validation scheme | Blocks | AUC | What it answers |
|---|---|---|---|
| random KFold | 854 | 0.927 ± 0.009 | nothing useful — see below |
| 0.1° blocks (~9 km) | 517 | 0.925 ± 0.023 | interpolate between neighbors |
| 0.25° blocks (~23 km) | 223 | 0.928 ± 0.022 | interpolate between neighbors |
| 0.5° blocks (~46 km) | 69 | 0.919 ± 0.049 | score the state with training points scattered nearby. This is how the map above is used |
| 1.0° blocks (~93 km) | 24 | 0.857 ± 0.043 | generalize to a region held out entirely |
| 2.0° blocks (~186 km) | 8 | 0.815 ± 0.049 | too few blocks to quote |
Baseline AUC 0.500; baseline average precision 0.404 (positive-class prevalence). 854 points, 345 positive.
Mineral occurrences are spatially autocorrelated. A point 200 m from a known mine sits in the same map unit. It carries the same lithology and the same age. Shuffle those into different folds and the held-out point is one the model has already seen, feature for feature, label included. Random cross-validation measures interpolation between neighbors. That is not the question being asked.
The gap between random and blocked validation is only 0.008 AUC at fine scales. That does not mean there is no leakage. It means the blocks are too small. Every feature is a map-unit attribute, so a 9 km block holds the same geology as its neighbors and the fold boundary cuts nothing. Only blocks large enough to hold out whole regions reveal the effect. At 186 km the score falls to 0.815.
Ranked by map unit, not by grid cell. Limitation 6 explains why.
| P | Cells | Centroid | Map unit |
|---|---|---|---|
| 0.920 | 58 | 34.10, -84.30 | Neoproterozoic-Cambrian sedimentary |
| 0.910 | 7 | 33.70, -82.60 | Cambrian volcanic: felsic rocks |
| 0.888 | 1 | 34.20, -84.00 | Mesoproterozoic crystalline metamorphic rocks |
| 0.796 | 1 | 33.90, -85.10 | Devonian volcanic: mafic rocks |
| 0.737 | 17 | 34.60, -84.50 | Neoproterozoic sedimentary |
| 0.737 | 37 | 34.70, -84.30 | Neoproterozoic sedimentary rocks |
| 0.465 | 2 | 33.80, -84.75 | Devonian plutonic: undivided granitic rocks |
| 0.451 | 162 | 33.70, -83.90 | Neoproterozoic-Cambrian volcanic: interlayered sedimentary and volcanic rocks |
lith_quartz fires on zero
rows of 854. lith_schist fires on one. Quartz-vein-in-schist
is the host rock for Georgia gold, and the model cannot see it.USGS MRDS occurrences and Macrostrat map units, joined by cached point lookups. Deduplicated to one row per location. Featurized into age and lithology features under an enforced allowlist. Trained with XGBoost, validated spatially. A 0.1° grid over the state runs through the same code path and gets scored.
The method, every design decision with what was rejected, and the steps to reproduce it are in the repository.