Exoplanet Detection ML: Detection of Exoplanets with Machine Learning Techniques through Transit Light Curve Analysis

This project is ongoing and subject to continuous advancements and modifications.

Exoplanet Detection ML is a machine learning project dedicated to the detection of exoplanets using transit survey-based light curves. By leveraging advanced machine learning algorithms and feature engineering techniques, this project aims to enhance the accuracy and efficiency of exoplanet discovery.

Table of Contents

• Features
• Machine Learning Algorithms
• Key Notebooks
• Examples
  • Model Performance
  • Light Curve Visualization
  • Workflow
• Project Structure
• Resources
  • Dimensionality Reduction
  • Feature Selection
  • Supervised Learning
  • Gaussian Process
  • Unsupervised Learning
  • Hyperparameter Tuning
  • Incremental PCA
  • Plotting
  • Probability Calibration
  • Additional Resources
• Acknowledgements
• License

Features

• Automated Exoplanet Detection: Utilizes transit survey-based light curves to identify potential exoplanets.
• Advanced Algorithms: Implements state-of-the-art machine learning models for high accuracy.
• Feature Engineering: Employs robust feature extraction and selection techniques to enhance model performance.
• Dimensionality Reduction: Reduces feature space complexity while preserving essential information.

Light Curve Visualization

Machine Learning Algorithms

Exoplanet ML employs a variety of machine learning algorithms to ensure comprehensive analysis and accurate predictions:

• Random Forest Classifier
• LightGBM
• AdaBoost
• Histogram Gradient Boosting
• XGBoost
• XGBoost Calibrated

Key Notebooks

• Kepler Lightcurve Notebook
• Feature Engineering with TSFresh
• Model Training and Evaluation

---

Workflow

---

Below are some examples of model performance:

Model Performance

Machine Learning Models	Accuracy	Precision	Sensitivity	F1-Score	ROC-AUC Score
Random Forest	84%	85%	84%	83%	85%
Adaptive Boosting	82%	82%	82%	80%	86%
Histogram Gradient Boosting	87%	87%	87%	87%	96%
Extreme Gradient Boosting	86%	87%	86%	85%	95%
Extreme Gradient Boosting (Calibrated)	89%	89%	89%	89%	93%

Confusion matrix

Resources

Dimensionality Reduction

• Introduction to PCA, t-SNE, and UMAP
• Plotly t-SNE and UMAP Projections
• Kernel PCA in scikit-learn
• Understanding UMAP
• UMAP Documentation

TsFresh Feature Selection

• TsFresh API Documentation
• TsFresh for Industrial Applications

Scikit-Learn Supervised Learning List and Description

• Scikit-Learn Supervised Learning

Gaussian Process

• Scikit-Learn Gaussian Process
• Gaussian Process Classifier
• Gaussian Process Kernels

Scikit-Learn Unsupervised Learning List and Description

• Scikit-Learn Unsupervised Learning

Hyperopt Hyperparameter Tuning

• Hyperopt Getting Started
• Hyperopt Tutorial on Kaggle

Incremental Principal Component Analysis

• Principal Component Analysis with Python
• Incremental PCA in scikit-learn

Scikit-Learn Plotting

• Scikit-Learn Display Object Visualization
• Confusion Matrix

Probability Calibration

• Probability Calibration in scikit-learn
• Calibrated Classifier

Technical Problem Solution and Miscellaneous Links

• Issue #1280 - YOLOv7
• Scikit Optimize Issues

Acknowledgements

• Feature Engineering with TSFresh
• Exoplanet Archive Acknowledgements
• Exoplanet Archive DOI
• Exoplanet Archive Table View
• Exoplanet Archive Table Redirect

License

This project is licensed under the CC-BY-4.0.

---

Note: The rest of the code and additional files can be found in the following repositories:

• Exoplanet ML GitLab
• New Dataset GitLab
• TSFresh Extract GitLab
• Data Extract GitLab
• ExoData New GitLab
• ExoData GitLab
• Exoplanet ML FP GitLab
• Exoplanet ML Kepler GitLab

Contact

For any inquiries or feedback, please contact:

Mir Sakhawat Hossain
📧 Email | 🔗 Website | 🐦 Google Scholar