ML Predictions
Four models working together — XGBoost severity classification, DistilBERT NLP, Prophet forecasting and HDBSCAN clustering. Every result is explainable.
Severity Classifier
XGBoost · 5-level risk rating
Predict the severity of a new in-progress crisis event and see exactly which features drove the decision (SHAP attribution).
NLP Report Classifier
DistilBERT · 8 crisis categories
Hotspot Clustering
HDBSCAN · crisis zones by density
Spatial hotspots currently detected across active crisis events, ranked by a composite risk score of density + average severity.
No hotspots yet.
Regional Forecast
Prophet time-series · predicted crisis events with 95% confidence intervals
Forecast Regions on Map
Predicted hotspot locations by density risk
XGBoost v1 · DistilBERT fine-tuned (fallback: keyword model) · Prophet (fallback: seasonal-naive) · HDBSCAN (fallback: DBSCAN)