🎯 Conformal Prediction for Housing Price Regression

Advanced uncertainty quantification using conformal prediction methods on California Housing dataset

🚀 Quick Start

git clone https://github.com/OJules/Conformal-Prediction-Housing.git
cd Conformal-Prediction-Housing
pip install -r requirements.txt
jupyter notebook Prediction_Conforme_Jules_Odje.ipynb

🎯 Project Overview

This project implements conformal prediction methods for regression tasks with a focus on uncertainty quantification. Using the California Housing dataset, we demonstrate how conformal prediction provides statistically valid prediction intervals.

🔑 Key Features

• ✅ Conformal Prediction Implementation from scratch
• ✅ Statistical Guarantees for prediction intervals
• ✅ Uncertainty Quantification with coverage analysis
• ✅ Real-world Dataset (California Housing)
• ✅ Complete Pipeline from data prep to evaluation

📊 Methodology

Conformal Prediction Framework

1. Split Conformal Prediction approach
2. Nonconformity Scores calculation
3. Prediction Intervals with guaranteed coverage
4. Performance Evaluation metrics

Implementation Highlights

class ConformePredictionProject:
    def __init__(self, alpha=0.1):
        # Alpha = significance level (10% for 90% coverage)
        self.alpha = alpha
        
    def prepare_data(self, test_size=0.2, calib_size=0.25):
        # Train/Calibration/Test split for conformal prediction
        
    def fit_conformal_predictor(self):
        # Implement conformal prediction algorithm

📈 Results & Performance

Coverage Analysis

• Target Coverage: 90% (α = 0.1)
• Empirical Coverage: [Results from your implementation]
• Average Interval Width: [Results from your implementation]

Key Findings

• Demonstration of distribution-free uncertainty quantification
• Comparison with standard regression confidence intervals
• Analysis of prediction interval efficiency

🛠️ Technical Stack

• Language: Python 3.8+
• ML Framework: scikit-learn
• Data Processing: pandas, numpy
• Visualization: matplotlib
• Environment: Jupyter Notebook

📁 Project Structure

Conformal-Prediction-Housing/
├── Prediction_Conforme_Jules_Odje.ipynb    # Main implementation
├── Prediction_Conforme_Project.pdf          # Detailed report
├── requirements.txt                          # Dependencies (to be added)
├── README.md                                # This file
└── results/                                 # Output plots & metrics (to be added)

🎓 Academic Context

This work is part of my Master's research in Statistics at Université de Neuchâtel and MIASHS/AI at Université Lyon 2, focusing on:

• Uncertainty quantification in machine learning
• Conformal prediction theory and applications
• Statistical guarantees for predictive models

Related Thesis Work:

• Master's MIASHS: "Conformal Risk Control for Semantic Uncertainty Quantification in CT"
• Master's Statistics: "Statistical Efficacy Study of Prevention Campaigns in Switzerland"

📚 References

• Vovk, V., Gammerman, A., & Shafer, G. (2005). Algorithmic Learning in a Random World
• Angelopoulos, A. N., & Bates, S. (2021). A gentle introduction to conformal prediction and distribution-free uncertainty quantification
• Lei, J., G'Sell, M., Rinaldo, A., Tibshirani, R. J., & Wasserman, L. (2018). Distribution-free predictive inference for regression

🤝 Contributing

Interested in conformal prediction? Feel free to:

• Open issues for questions or suggestions
• Submit pull requests for improvements
• Connect with me for research collaboration

📫 Contact

Jules Odje - Data Scientist | Aspiring PhD Researcher
📧 odjejulesgeraud@gmail.com
🔗 LinkedIn
🐙 GitHub

---

"Providing statistical guarantees for uncertainty quantification in machine learning" ✨