bhargavvz/Aquasafe

💧 AquaHealth — Analysis and Prediction of Waterborne Health Diseases

A full-stack mobile health platform that uses gradient-boosted ensemble models to predict waterborne diseases, assess water quality, forecast outbreak risk, and evaluate sanitation conditions across India.

---

Repository Structure

├── app/            Flutter mobile application (cross-platform)
├── backend/        Flask REST API with JWT auth & ML inference
├── ML/             Model training pipeline, dataset, and plots
├── doc/            IEEE research paper (LaTeX)
└── README.md       ← you are here

Each subdirectory has its own README.md with setup instructions.

---

What AquaHealth Does

Model	Task	Algorithm	Key Metric
Disease Classifier	8-class classification (7 diseases + No Disease)	XGBoost + LightGBM weighted ensemble	95.19% accuracy
WQI Predictor	Water Quality Index regression	XGBoost Regressor	R² = 0.886
Outbreak Risk	Binary outbreak prediction	LightGBM	AUC-ROC = 0.9997
Water Safety	3-class water safety (Safe / Moderate / Dangerous)	XGBoost	99.25% accuracy
Sanitation Risk	Sanitation risk score regression	XGBoost Regressor	R² = 0.787

Diseases Covered

Cholera · Typhoid · Dysentery · Giardiasis · Hepatitis A · Hepatitis E · Leptospirosis

---

Architecture Overview

┌─────────────────────┐
│   Flutter Mobile App │  ← User-facing UI (19 screens)
│   (Dart / Riverpod)  │
└────────┬────────────┘
         │ HTTP + JWT
┌────────▼────────────┐
│   Flask REST API     │  ← Authentication, validation, inference
│   (Python / Gunicorn)│
└────────┬────────────┘
         │
┌────────▼────────────┐
│   ML Models (joblib) │  ← 5 pre-trained models loaded at startup
│   PostgreSQL (users) │
└─────────────────────┘

---

Quick Start

1. Train the models (or use pre-trained)

cd ML
pip install -r requirements.txt
python train_models.py

2. Start the backend

cd backend
pip install -r requirements.txt
# Set DATABASE_URL or use the default PostgreSQL connection
python app.py

3. Run the mobile app

cd app
flutter pub get
flutter run

Update the API base URL in app/lib/core/api_service.dart to point to your running backend.

---

Dataset

5.25 million synthetic records calibrated against real-world Indian epidemiological data (CPCB, IDSP, Census of India). Covers all 36 states/UTs, 615+ districts, 40 features spanning geography, water chemistry, sanitation, climate, demographics, and clinical symptoms.

See ML/Dataset-info.md for the complete data dictionary.

---

Tech Stack

Layer	Technology
Mobile App	Flutter 3.9+, Dart, Riverpod, fl_chart, flutter_map
Backend API	Python 3.10+, Flask, Gunicorn, PostgreSQL, PyJWT
ML Training	XGBoost, LightGBM, scikit-learn, pandas, matplotlib
Model Serialization	joblib
Authentication	JWT (7-day expiry), salted SHA-256 password hashing

---

Research Paper

The accompanying IEEE-format paper is in doc/research_paper.tex. It documents the methodology, dataset construction, model architecture, and experimental results.

---

License

This project is developed as an academic major project. See individual component READMEs for details.