madhavisolanki-ui/Aquanga

Ganga River water-quality forecasting and early-warning system using ML, deep learning, and geospatial analytics. # 🌊 Aquanga – Predictive Water Monitoring & Early Warning System **Aquanga** is a production-grade, end-to-end water quality forecasting and environmental early warning system for Central Pollution Control Board (CPCB) monitoring stations along the Ganga River basin. It integrates machine learning and deep learning time-series forecasting, real-time risk assessment, a RESTful FastAPI backend, a PostgreSQL relational datastore, and an interactive Streamlit geospatial dashboard. --- ## 📌 Table of Contents 1. Problem Statement 2. System Architecture 3. Dataset & Preprocessing 4. Feature Engineering 5. Machine Learning & Deep Learning Models 6. Model Evaluation & Benchmark 7. Environmental Risk & Early Warning System 8. FastAPI REST API 9. Database Architecture & Seeding 10. Interactive Streamlit Dashboard 11. Project Structure 12. How to Run Locally 13. Docker & Containerized Deployment 14. Testing Suite 15. Generalization & Future Improvements --- ## 🎯 Problem Statement The Ganga River supports over 400 million people but faces severe ecological stress from municipal sewage, industrial effluents, and agricultural runoff. Traditional water quality monitoring relies on post-hoc manual laboratory testing, often detecting hypoxic events (low Dissolved Oxygen) and severe microbial surges days after they occur. **Aquanga** solves this by: - Forecasting future **Dissolved Oxygen (DO)** levels using historical multi-parameter time-series observations. - Calculating environmental risk scores based on statutory CPCB water quality criteria. - Automatically generating actionable early warnings to alert environmental authorities before ecological thresholds are breached. --- ## 🏛️ System Architecture ```mermaid flowchart TD A[Raw CPCB Water Quality Data] --> B[Data Preprocessing & Imputation] B --> C[Feature Engineering & Compliance Flags] C --> D[Chronological Train / Test Split] D --> E[6 ML & DL Models Training] E --> F[Evaluation Benchmark & Best Model Selectio …