Apppublic

ZeniTsuIsSimP/Water-Contamination-Detection

sourceHugging Faceupdated 4mo agoView on Hugging Face

0likes

App README

💧 Smart Water Contamination Detection

AI-powered water quality monitoring system using IoT sensor data, Machine Learning, Docker, and CI/CD — built with a professional MLOps pipeline.

🔗 Live Project

—Hugging Face Spaces: https://huggingface.co/spaces/ZeniTsuisSimp/Water-Contamination-Detection

🏗️ Project Structure

├── app/                    # Streamlit web application
│   └── main.py
├── src/                    # ML pipeline source code
│   ├── data_preprocessing.py   # Data loading, imputation, splitting
│   ├── train.py                # Model training + MLflow logging
│   ├── evaluate.py             # Evaluation + report generation
│   └── predict.py              # Prediction + anomaly detection
├── data/
│   ├── raw/                # Original, unprocessed CSVs
│   └── processed/          # Classified & cleaned datasets
├── notebooks/
│   └── model_training.ipynb    # Exploratory notebook (EDA + training)
├── tests/
│   └── validate_model.py       # CI model validation test
├── reports/                # Auto-generated metrics, plots
├── mlruns/                 # MLflow experiment tracking data
├── params.yaml             # Centralized hyperparameters & config
├── dvc.yaml                # DVC pipeline definition
├── Dockerfile              # Docker containerization
├── docker-compose.yml      # Docker Compose for local dev
├── requirements.txt        # Python dependencies
└── .github/workflows/
    └── ci-cd.yml           # GitHub Actions CI/CD pipeline

🚀 Quick Start

1. Setup

bash

# Clone the repository
git clone <repo-url>
cd Water-Contamination-Detection

# Create virtual environment
python -m venv venv
venv\Scripts\activate        # Windows
# source venv/bin/activate   # Linux/Mac

# Install dependencies
pip install -r requirements.txt

2. Train the Model

bash

python -m src.train

This trains the model, logs to MLflow, and saves water_model.pkl.

3. Evaluate

bash

python -m src.evaluate

Generates classification reports and confusion matrix plots in reports/.

4. Run the App

bash

streamlit run app/main.py

Open http://localhost:8501 in your browser.

5. Run with Docker

bash

docker-compose up --build

🔬 ML Pipeline

Stage	Command	Output
Preprocess	`python -m src.data_preprocessing`	`reports/data_summary.json`
Train	`python -m src.train`	`water_model.pkl`, `reports/metrics.json`
Evaluate	`python -m src.evaluate`	`reports/eval_metrics.json`, `reports/confusion_matrix.png`

DVC Pipeline

bash

dvc repro        # Run full pipeline
dvc status       # Check if pipeline is up-to-date
dvc dag          # Visualize pipeline DAG

MLflow

bash

mlflow ui        # Open experiment tracker at http://localhost:5000

🔧 Configuration

All hyperparameters and thresholds are in `params.yaml`:

—Data paths — raw/processed data locations
—Feature names — pH, Solids (TDS)
—Model hyperparameters — RF trees
—Anomaly thresholds — critical pH/TDS limits

🧠 Hybrid Prediction Logic

1.Rule-Based Safety Check — Catches critical anomalies (pH < 4 or > 10, TDS > 3000)
2.ML Model — Random Forest predicts potability for non-critical inputs
3.Confidence Smoothing — Prevents unrealistic 100%/0% probabilities

🚢 CI/CD Pipeline

Stage	Description
Lint & Test	Flake8 linting + model validation test
Docker Build	Build image + health check
Deploy	Push to Hugging Face Spaces (main branch only)

📊 Tech Stack

—ML: scikit-learn (Random Forest)
—App: Streamlit + Altair charts
—Experiment Tracking: MLflow
—Data Versioning: DVC
—Containerization: Docker
—CI/CD: GitHub Actions → Hugging Face Spaces