samabesherr-droid/Aether

AI-Powered Environmental Risk Monitoring & Forecasting System for Cairo, Egypt **Predicting environmental health risks in Cairo through data engineering and machine learning** Overview · Architecture · Quick Start · ML Model · Dashboard · Contributing --- ## Overview Aether is a production-grade environmental health intelligence platform built for Cairo, Egypt. It ingests real-time weather and air quality data from live APIs, processes it through an automated ETL pipeline, and applies a trained Random Forest classifier to generate daily health risk predictions — delivering actionable insights to residents through an interactive Power BI dashboard. The platform was built to address a genuine public health gap: Cairo consistently ranks among the world's most polluted cities, yet there is no unified, automated system that translates raw environmental data into clear, daily health guidance. ``` 📡 Live APIs → 🔧 ETL Pipeline → 🗄️ SQL Server DW → 🤖 ML Model → 📊 Power BI (Prefect orchestrates every step, daily) ``` **Key outcomes:** - Classifies each day into one of 5 health risk categories with 97% F1 score - Tracks 13 environmental features across weather and pollution dimensions - Covers 1,154 days of historical data (2022–2025) for trend analysis - Fully automated — runs daily with zero manual intervention --- ## Architecture The platform is organised into four clearly separated layers: **Ingestion layer** — Two Python scripts pull from OpenWeatherMap (weather) and WAQI (air quality, station A527650 Cairo). Historical backfill CSVs cover 2000–2026. **Processing layer** — `data_cleaning.py` applies targeted null handling (median fill on non-critical columns, drop only when critical fields are missing). `feature_engineering.py` inner-joins weather and pollution on `date` and derives five composite health features. **Storage layer** — SSIS packages load `environmental_features.csv` into SQL Server (`AetherDW_V0`) with five tables. A lookup transformation prevents duplicate loads. **Intelligence layer** — A Random Forest classifi …