Jcarljuson/ExoFinder-by-Aethereologists

🪐 ExoFinder by Aethereologists

Advanced Machine Learning Exoplanet Detection System

ExoFinder is a sophisticated web application that uses ensemble machine learning models to analyze astronomical data and predict the likelihood of exoplanet candidates. Built by the Aethereologists team for NASA Space Apps Challenge.

🚀 Features

• —Multi-Mission Support: Analyze data from Kepler, K2, and TESS missions
• —Ensemble ML Models: Uses Random Forest, XGBoost, and Neural Networks
• —Real-time Analysis: Upload CSV data and get instant predictions
• —Interactive Interface: User-friendly web interface with space-themed design
• —Confidence Scoring: Get probability scores for exoplanet likelihood

🔬 Technical Details

Machine Learning Models

• —Kepler Mission: Stacked ensemble model (Random Forest + XGBoost + Neural Network)
• —K2 Mission: Stacked ensemble model with specialized preprocessing
• —TESS Mission: Optimized model with label encoding for categorical features

Key Features Analyzed

• —Transit depth and duration
• —Orbital period and characteristics
• —Stellar parameters (temperature, radius, magnitude)
• —Signal-to-noise ratios
• —Centroid motion and flux measurements

📊 Usage

1. 1.Select your mission type (Kepler, K2, or TESS)
2. 2.Upload your CSV file with astronomical measurements
3. 3.Click "Analyze Data" to run the ML prediction
4. 4.View results with confidence scores and detailed analysis

🛠️ About

Developed using Flask, scikit-learn, XGBoost, and TensorFlow. The models are trained on validated exoplanet datasets and achieve high accuracy in distinguishing between confirmed exoplanets and false positives.