How It Works

How HantaMap.io collects, classifies, forecasts, and publishes hantavirus outbreak intelligence worldwide.

Data Sources

HantaMap monitors 185+ RSS sources across 41 countries in real-time, spanning health authorities, international organisations, and local media.

• Health authorities — WHO, CDC, ECDC, PAHO, RKI, THL, FHI, KDCA, national ministries of health, and more.
• International aggregators — ProMED-mail, ReliefWeb, CIDRAP.
• National media — BBC, NHK, Tagesschau, France Info, ANSA, El País, ABC News AU, and 100+ regional outlets.
• GDELT 2.0 DOC API — global news monitoring with relevance filtering (every 30 min).

Sources are validated weekly via automated probes. Broken feeds are flagged and visible in the admin Source Coverage panel.

Map Layers

The HantaMap displays six distinct, independently togglable layers:

1. Verified Outbreaks

Confirmed cases with location, case counts, and dates verified by an official health authority. Displayed as solid markers with status badges (Confirmed, Monitoring, Suspected, Closed).

2. Correlated Signals

Multiple independent sources reporting consistent evidence of hantavirus activity in the same region. These are not yet officially confirmed but show strong corroboration.

3. Early Signals

First-time mentions or single-source reports that may indicate emerging activity. These enter the analyst review queue for validation.

4. Signal Activity / Mentions

Aggregated mention volume from RSS feeds and GDELT, shown as heat or density overlays. Useful for spotting regions with growing media attention.

5. Endemic Zones

Known regions with established hantavirus reservoir populations and historical transmission cycles. Displayed as shaded polygons.

6. Predictive Risk

Forward-looking risk forecast generated by the HantaRisk Forecast Model (v1). Estimates where hantavirus risk may increase in the next 7–30 days.

• Risk levels: Low · Moderate · High · Very High
• Drivers: signal activity growth, source diversity, endemic zone proximity, seasonality, recent case mentions, climate proxy
• Displayed as pulsing gradient circles with colour-coded risk levels
• Each zone includes an explanation of the top contributing factors

⚠️ Important: Predictive risk is an analytical forecast, not a confirmed outbreak or medical advice. It is clearly separated from verified outbreak data.

Confidence Scoring

Every outbreak record carries a confidence score from 0 to 1, derived from: source credibility weight, AI classification certainty, corroboration across independent sources, and analyst review status. Only signals ≥ 0.85 from trusted sources are auto-promoted; everything else enters the human review queue.

Source Validation

Every RSS source is assigned a credibility score (0–100) based on institutional authority, editorial standards, and track record. Sources are categorised as health_authority, international_org, or local_media.

• Health authority and international org sources can auto-promote high-confidence signals.
• Local media sources only provide signals for analyst review — they cannot auto-publish.
• Automated weekly validation probes check feed availability and RSS structure.
• Broken or inactive feeds are flagged with diagnostic errors and timestamps.

How Source Credibility Is Calculated

Each source carries a credibility_score from 0 to 1 (shown as 0–100% on the Sources page). It is the strongest input to every signal's overall confidence.

Initial assignment

Scores are seeded by institutional authority and editorial track record:

• Health authorities & international orgs (WHO, CDC, ECDC, ProMED): 0.90 – 1.00, marked trusted.
• Tier-1 international wires (Reuters, AP, BBC): 0.75 – 0.85.
• National & regional media: ~0.55, not trusted by default.
• Blogs, social, unknown: 0.10 – 0.30.

Resolution at classification time

For each ingested document, the classifier picks the source trust value in this order:

1. Database credibility_score on the source (analyst-curated, highest priority).
2. Domain reputation map (who.int, cdc.gov, reuters.com, …).
3. The AI's own source_credibility label (official, scientific, news, local_media, blog, social) mapped to a trust value.

Weight in signal confidence

Source trust accounts for 40% of every signal's final confidence:

finalConfidence = sourceTrust       × 0.40
                + aiConfidence      × 0.20
                + geoCompleteness   × 0.15
                + dataCompleteness  × 0.25

• Signals below 0.40 are auto-rejected.
• Signals at ≥ 0.85 from official or scientific sources can be auto-published.
• Everything else enters the analyst review queue.

Note: credibility is currently set by editorial policy and admin review. An automated feedback loop that adjusts scores based on a source's historical accuracy (rejected vs. confirmed signals) is planned but not yet active.

Geographic Coverage

Priority coverage across hantavirus-relevant regions:

Mobile & API Access

Predictive risk data is available via a public API endpoint for mobile and third-party integrations, returning nearby risk zones, scores, top drivers, and disclaimers.

Limitations

• Open-source signals cannot replace official surveillance systems.
• Geolocation may be imprecise at city level.
• Case counts evolve — records reflect best-known values at last update.
• Predictive risk is rule-based (v1) and does not yet incorporate machine-learning models or real-time climate data feeds.
• Some countries have fewer than 3 active sources — coverage gaps are tracked in the admin panel.

Disclaimer

HantaMap.io is an outbreak intelligence and epidemiological monitoring platform. It does not replace official public health guidance. Predictive risk layers are analytical forecasts, not confirmed outbreaks. Always refer to WHO, CDC, ECDC, and national health authorities for official recommendations.