Methodology
How signals become operational intelligence
GlobalSystem AI normalizes source feeds into a consistent event model so each module can compare events across source, severity, time, geography, and confidence.
Normalization
Incoming provider payloads are transformed into a shared event contract: id, type, title, category, coordinates, severity, timestamp, source, country approximation, source link, and confidence score.
Severity
Severity is normalized per signal type. Earthquakes use magnitude bands, while weather and event categories use provider signal quality and recency until deeper module-specific models are added.
Freshness and cache behavior
External feeds are cached to reduce upstream load and keep dashboards resilient. When a live request fails, cached data can continue powering the interface and is marked as stale in API payloads.
Current limitations
Country enrichment is an approximate bounding-box model until a production geocoder or administrative boundary dataset is integrated. Confidence and severity scoring should be treated as operational indicators, not certified risk ratings.