AI-Powered Exoplanet Classification using NASA's Kepler, K2, and TESS Mission Data
An intelligent machine learning system that classifies celestial objects as Confirmed Exoplanets, Planetary Candidates, or False Positives using NASA's open-source data. Built for the NASA Space Apps Challenge 2024, this tool brings exoplanet discovery capabilities to researchers and astronomy enthusiasts worldwide.
- AI-Powered Classification: 75.7% accurate XGBoost model trained on real NASA data
- Automated Feature Engineering: Converts 10 basic astronomical measurements into 31 scientific features
- User-Friendly Interface: Clean Streamlit web app for both single and batch predictions
- Real NASA Data: Trained on combined datasets from Kepler, K2, and TESS missions
- Professional Results: Confidence scores and detailed classification reports
| Metric | Score |
|---|---|
| Validation Accuracy | 75.7% |
| Balanced Accuracy | 75.0% |
| F1-Score (Macro) | 75.5% |
| Best Class (False Positive) | 77.3% |
- Access the Web App: Live Demo
- Single Prediction: Enter 10 basic astronomical parameters
- Batch Analysis: Upload CSV files for multiple candidates
- Get Results: Instant classifications with confidence scores
10 Base Parameters Required:
- Orbital Period (days)
- Transit Duration (hours)
- Transit Depth (ppm)
- Planet Radius (Earth radii)
- Semi-major Axis (AU)
- Impact Parameter
- Star Temperature (K)
- Star Radius (Solar radii)
- Star Gravity (log g)
- Signal-to-Noise Ratio
# Clone repository
git clone https://github.com/yourusername/nasa-exoplanet-hunter.git
# Install dependencies
pip install -r requirements.txt
# Run locally
streamlit run app.py