ngohjuniormbah/aeroshield-cameroon

# AeroShield Cameroon **AeroShield Cameroon** is a full-stack AI project designed for the **IndabaX Cameroon 2026 Hackathon**. It turns the provided meteorological dataset into a **virtual air-quality sensing system** that forecasts **next-day air pollution risk** for cities across Cameroon, then exposes the results through a **FastAPI backend** and a **Streamlit dashboard**. ## Why this project can stand out The official baseline dataset does not include a measured PM2.5 column. Instead of stopping there, this project introduces a strong competition strategy: - build a **transparent weather-to-air-quality proxy** - forecast **next-day city-level risk**, not just same-day scoring - create a **national virtual sensor network** without expensive hardware - provide a **decision dashboard** with a heatmap, climate comparison, and alert levels This fits the hackathon brief while also telling a stronger story: **Cameroon can monitor pollution risk at scale using climate data and AI, even where physical sensors are scarce.** ## Project structure ```text AeroShield Cameroon/ ├── api/ │ └── main.py ├── core/ │ ├── feature_engineering.py │ ├── train_pipeline.py │ └── xlsx_loader.py ├── dashboard/ │ └── app.py ├── docs/ │ ├── JUDGES_GUIDE.md │ └── SUBMISSION_GUIDE.md ├── artifacts/ ├── requirements.txt └── run.sh ``` ## Core idea ### 1. Virtual air-quality layer The dataset contains weather variables such as temperature, rainfall, wind, solar radiation, evapotranspiration, city, region, latitude, and longitude. AeroShield converts those signals into an **AQRI**: **AQRI = Air Quality Risk Index** The AQRI is a transparent risk score influenced by: - low wind - low rainfall - heat - strong solar radiation - dryness - stagnation or dust-related weather conditions ### 2. Forecasting task The model uses current and lagged weather conditions to forecast: - **next-day AQRI score** - **next-day alert level**: Low, Medium, High ### 3. User-facing system Th …