Combines XGBoost impact regression with pgvector semantic search over 2,281 historical disaster reports to simulate affected population and economic damage for any hazard event, anywhere in the world.
01
Configure the Event
Select a country, hazard class (earthquake, flood, etc.), month, year, and severity on the Richter or Saffir–Simpson scale.
02
Run the Math Engine
XGBoost regression models — trained separately for each disaster type — predict estimated affected population and economic impact in USD.
03
Retrieve Historical Context
pgvector performs a cosine similarity search over 2,281 embedded disaster narratives to surface the 3 most analogous historical events by meaning.