m-eze/naija-security-forecast

LGA-level security risk forecast for Nigeria — FastAPI + PostGIS backend, Next.js choropleth frontend # 🇳🇬 Naija Security Forecast A "weather forecast for security" — a live, LGA-level risk dashboard for Nigeria that fuses historical incident data, news sentiment, and statistical trend models into a choropleth map anyone can read at a glance. --- ## What it does | Feature | Detail | |---|---| | **Choropleth map** | All 775 Nigerian LGAs coloured LOW → MODERATE → HIGH → SEVERE | | **Timeline slider** | Slide from −3 days (historical) through today to +3 days (forecast); scores use trend velocity + news sentiment decay | | **Incident type filters** | Kidnappings, school abductions, jihadist attacks, bombings, banditry, riots, protests, and more | | **Live news feed** | Scrapes 10 Nigerian RSS sources; NLP classifies security relevance and sentiment | | **State rankings** | Left sidebar ranks all 36 states + FCT by average risk score | | **LGA detail panel** | Click any LGA on the map for a full score breakdown | --- ## Architecture ``` naija-security-forecast/ ├── app/ │ ├── api/routes/ # FastAPI routers (risk, news, incidents, filters) │ ├── core/ # DB engine, settings │ ├── models/ # SQLAlchemy ORM (LGA, Incident, NewsArticle, RiskScore) │ ├── schemas/ # Pydantic v2 response schemas │ ├── services/ │ │ ├── acled_client.py # ACLED API client (email + access_key auth) │ │ ├── acled_sync.py # Historical incident sync (from 2010) │ │ ├── forecaster.py # 7-day statistical forecast engine │ │ ├── lga_matcher.py # Fuzzy LGA name → PostGIS point-in-polygon matcher │ │ ├── news_pipeline.py # Scrape → NLP → upsert pipeline │ │ ├── nlp.py # Keyword NLP: security classification + sentiment │ │ ├── rss_scraper.py # feedparser scraper for 10 Nigerian news sources │ │ └── scorer.py # Composite risk scoring engine │ └── workers/ # Celery tasks + beat schedule ├── alembic/ # DB migrations ├── scripts/ # Manual t …