Daily ML-driven volatility predictions for 12 exchange-traded minerals across 7, 14, and 30-day horizons. One structured dataset, delivered before market open.
Background colour shows the predicted risk level each day. Arrows mark days where realised volatility exceeded the 1.5σ threshold — the events the model is trained to detect.
We predict the probability of abnormal price volatility — not whether prices will rise or fall. This is a deliberate design choice: for most institutional use cases, the magnitude of uncertainty is more actionable than a directional guess.
A HIGH risk level for copper means elevated probability of a large price move in either direction within the forecast window. That single signal informs options pricing, position sizing, hedge timing, and risk limits — without requiring a view on direction.
Five tiers from LOW to EXTREME, calibrated to the probability of volatility exceeding the trailing 60-day baseline + 1.5σ.
Boolean field for automated monitoring. True only when HIGH or EXTREME with sufficient confidence. Pipe directly into alerting systems.
Three named drivers per prediction across 16 signal categories — news volume, production risk, geographic spread, and more.
Transparent data density indicator. Know whether an assessment rests on 10+ events or sparse signals and momentum alone.
One day's predictions across all minerals. This is exactly what arrives in your pipeline each morning.
| Mineral | Exchange | Risk Level | Alert | Confidence | Factor 1 | Factor 2 | Factor 3 |
|---|---|---|---|---|---|---|---|
| Copper | LME | HIGH | ● ALERT | high | news volume 7d | production risk 7d | country concentration 7d |
| Lithium | LME | EXTREME | ● ALERT | high | battery supply chain 7d | news acceleration | country concentration 7d |
| Cobalt | LME | HIGH | ● ALERT | medium | production risk 7d | country concentration 7d | high severity events 7d |
| Nickel | LME | MODERATE | — | medium | news volume 7d | price return 5d | historical volatility 60d |
| Gold | COMEX | LOW | — | high | price vs 60d average | historical volatility 60d | news volume 7d |
| Aluminium | LME | ELEVATED | — | high | news volume 7d | event magnitude 7d | media coverage 7d |
| Zinc | LME | LOW | — | low | historical volatility 60d | price vs 20d average | news volume 7d |
| Tin | LME | ELEVATED | — | medium | country concentration 7d | production risk 7d | supply concentration 7d |
| Silver | COMEX | MODERATE | — | medium | cross mineral contagion | price return 5d | news volume 7d |
| Palladium | NYMEX | HIGH | ● ALERT | high | production risk 7d | country concentration 7d | high severity events 7d |
| Platinum Group | NYMEX | ELEVATED | — | medium | production risk 7d | news volume 7d | country concentration 7d |
| Iron Ore | SGX | MODERATE | — | high | news volume 7d | event magnitude 7d | price vs 20d average |
| Lead | LME | LOW | — | low | historical volatility 60d | price return 5d | news volume 7d |
Sample: 2026-03-10 · 13 rows · 7d horizon
Probability of realised volatility exceeding trailing 60-day average + 1.5 standard deviations within forecast horizon
The top 3 risk factors are named in every row. 16 signal categories across 50+ individual features.
GDELT Global Knowledge Graph — 96 files per day covering 100+ languages. Every article mentioning critical mineral supply disruptions is captured, classified, and severity-scored.
136 features per mineral per day across rolling windows (3, 7, 14, 30 days). Includes news volume, severity analytics, geographic spread, supply concentration, price momentum, and cross-mineral contagion signals.
XGBoost classifiers with walk-forward cross-validation on 14+ months of historical data. Models retrained monthly. Separate model per mineral per forecast horizon.
Mean backtested AUC of 0.815 across all active models. Models below 0.65 AUC are excluded. Currently 34 of 36 possible models meet the quality threshold.
Probability of abnormal volatility — defined as realised volatility exceeding the trailing 60-day average plus 1.5 standard deviations within the forecast window. Captures large moves in either direction.
Every post is generated from live pipeline data — real scores, real movements, real disruption events. Not generic commentary.
Chile's copper production signals have deteriorated sharply over the past week, triggering a HIGH risk classification across our 7-day forecast horizon.
Platinum group metals maintained elevated risk levels throughout January 2025, driven by persistent South African supply disruptions and geopolitical tensions.
A retrospective analysis of how our ML models detected early signals of the July 2023 gallium export restrictions and the cascade effect on related minerals.
An explanation of how the country concentration feature quantifies geographic supply risk using Herfindahl-Hirschman Index calculations across producing nations.
When lithium volatility spikes, cobalt and nickel often follow within 1-3 trading days. Our cross-mineral contagion feature captures this relationship.
Posts are generated from live pipeline data using the same GDELT signals and ML features that power the predictions. Published weekly across five rotating content categories.
One dataset, delivered before market open.
Models are trained primarily on English-language news sources. Coverage of Chinese domestic policy signals is limited. Performance during unprecedented systemic crises may differ from backtested results. Risk levels indicate probability of abnormal volatility, not price direction. This product should be used as one input among many in investment decisions — not as a standalone trading signal.