Water Quality Prediction System

A machine learning system for predicting water quality based on physicochemical parameters. This project uses various ML algorithms to classify water quality into three categories: Excellent, Good, and Poor.

📁 Project Structure

Water clarity/
├── README.md                   # This file
├── Water-Clarity-DS.csv       # Dataset containing water quality measurements
├── test.ipynb                 # Jupyter notebook for model development and training
├── water_quality_model.pkl    # Trained machine learning model (serialized)
├── feature_names.json         # List of feature names used by the model
├── class_labels.json          # Mapping of class indices to quality labels
├── predict.py                 # Standalone prediction script
├── api.py                     # API endpoint for model serving
└── __pycache__/              # Python cache files
    └── api.cpython-311.pyc

🔬 Features

The model uses the following water quality parameters to make predictions:

• Temperature (°C)
• Turbidity (cm)
• Dissolved Oxygen (mg/L)
• BOD (Biological Oxygen Demand, mg/L)
• pH value
• Ammonia concentration (mg/L)
• Nitrite concentration (mg/L)

🎯 Model Output

The system predicts water quality in three categories:

• 0: Excellent water quality
• 1: Good water quality
• 2: Poor water quality

🚀 Getting Started

Prerequisites

pip install pandas numpy scikit-learn joblib matplotlib seaborn

Quick Start

1. Using the prediction script directly:

from predict import predict_water_quality

# Example prediction
result = predict_water_quality(
    temp=67.45,
    turbidity=10.13,
    do=0.208,
    bod=7.474,
    ph=4.752,
    ammonia=0.286,
    nitrite=4.355
)

print(f"Water Quality: {result['quality']}")
print(f"Confidence: {result['probabilities']}")

2. Running the API server:

python api.py

3. Training your own model:
   • Open test.ipynb in Jupyter Notebook
   • Run all cells to train and evaluate different models
   • The best model will be automatically saved

📊 Dataset Information

The dataset (Water-Clarity-DS.csv) contains water quality measurements with:

• European decimal format (comma-separated)
• Multiple physicochemical parameters
• Balanced sampling across quality categories

🤖 Model Development

Available Models

The system evaluates multiple machine learning algorithms:

• Random Forest
• Gradient Boosting
• Support Vector Machine (SVM)
• Logistic Regression
• K-Nearest Neighbors

Model Selection Process

1. Data Preprocessing: Handle European decimal format, balance classes
2. Model Evaluation: Cross-validation with stratified sampling
3. Hyperparameter Tuning: Grid search for optimal parameters
4. Performance Analysis: Classification reports and confusion matrices

Key Functions in test.ipynb

• load_and_preprocess_data(): Load and clean the dataset
• balance_classes(): Balance the dataset for fair training
• evaluate_models(): Compare different ML algorithms
• optimize_best_model(): Hyperparameter tuning for best model
• analyze_feature_importance(): Understand feature contributions

📈 Model Performance

The system automatically selects the best performing model based on cross-validation scores. Typical performance metrics include:

• Accuracy scores for each class
• F1-scores per quality category
• Confusion matrix analysis
• Feature importance rankings

🔧 API Usage

The API provides a REST endpoint for making predictions:

# Example API call structure
POST /predict
{
    "temp": 67.45,
    "turbidity": 10.13,
    "do": 0.208,
    "bod": 7.474,
    "ph": 4.752,
    "ammonia": 0.286,
    "nitrite": 4.355
}

📝 Model Artifacts

The system generates several important files:

• water_quality_model.pkl: Serialized trained model
• feature_names.json: Feature names in correct order
• class_labels.json: Quality label mappings

🔍 Feature Importance

The model analyzes which parameters are most important for water quality prediction. This helps understand:

• Which measurements have the strongest impact on water quality
• How different parameters contribute to the final classification
• Insights for water quality monitoring priorities

🎨 Visualization

The Jupyter notebook includes:

• Data distribution plots
• Model performance comparisons
• Feature importance visualizations
• Confusion matrix heatmaps

📋 Usage Examples

1. Batch Predictions

import pandas as pd
from predict import predict_water_quality

# Load your data
data = pd.read_csv('new_water_samples.csv')

# Make predictions for each row
predictions = []
for _, row in data.iterrows():
    result = predict_water_quality(
        temp=row['temp'],
        turbidity=row['turbidity'],
        do=row['do'],
        bod=row['bod'],
        ph=row['ph'],
        ammonia=row['ammonia'],
        nitrite=row['nitrite']
    )
    predictions.append(result['quality'])

data['predicted_quality'] = predictions

2. Real-time Monitoring

# Integrate with sensor data
def monitor_water_quality(sensor_data):
    result = predict_water_quality(**sensor_data)

    if result['prediction'] == 2:  # Poor quality
        send_alert(f"Poor water quality detected: {result['quality']}")

    return result

🛠️ Development

Extending the Model

1. Add new features to the dataset
2. Update feature_names.json accordingly
3. Retrain the model using test.ipynb
4. Test with new prediction scripts

Improving Performance

1. Collect more training data
2. Experiment with feature engineering
3. Try advanced algorithms (XGBoost, Neural Networks)
4. Implement ensemble methods

📊 Model Evaluation Metrics

The system provides comprehensive evaluation:

• Cross-validation scores: Generalization performance
• Test accuracy: Final model performance
• Per-class metrics: Precision, recall, F1-score
• Confusion matrix: Classification details

🔄 Retraining

To retrain the model with new data:

1. Update Water-Clarity-DS.csv with new samples
2. Run the complete pipeline in test.ipynb
3. New model artifacts will be automatically saved
4. Update prediction scripts if needed

💡 Best Practices

1. Data Quality: Ensure measurements are accurate and consistent
2. Feature Scaling: The system handles scaling automatically
3. Class Balance: Dataset balancing is implemented for fair training
4. Validation: Always validate predictions with ground truth when possible

🤝 Contributing

1. Fork the repository
2. Create your feature branch
3. Add tests for new functionality
4. Update documentation
5. Submit a pull request

📄 License

This project is open source and available under the MIT License.

🆘 Support

For issues or questions:

1. Check the Jupyter notebook for detailed examples
2. Review the prediction script for usage patterns
3. Examine the API code for integration examples

---

Note: This system is designed for educational and research purposes. For production water quality monitoring, please validate against certified laboratory measurements and follow local regulations.