Solar Flare Prediction – RHESSI Mission

Overview

This project focuses on predicting solar flare energy ranges using machine learning techniques, developed for the NASA Space Apps Challenge 2024 by Team Arjuna. Solar flares can have significant impacts on Earth's space environment, and accurate predictions are crucial for mitigating their effects.

For more information about our team, visit our page here.

Data Preprocessing

The dataset used in this project was obtained from the NASA RHESSI Data Repository and underwent several preprocessing steps to prepare it for analysis and modeling. The key steps include:

1. Data Source: The dataset was sourced from the NASA RHESSI Data Repository, accessible via a web form on this page.

2. FITS to CSV Conversion: The original data files were in FITS format, which were converted to CSV format using tools available here.

3. Dataset Compilation: Multiple CSV files generated from the FITS files were concatenated to create the current dataset. The script used to perform this concatenation can be found in this repository.

The complete dataset is available in the repository here.

Project Breakdown

Analysis

The data analysis phase involved:

• Understanding the characteristics of solar flares, including their energy distribution, duration, and occurrence over time.
• Identifying patterns in the dataset to guide the feature selection for predictive modeling.

Prediction

Various machine learning models were built and trained to predict the energy range of a solar flare based on its attributes. The models include:

• Linear Regression
• Logistic Regression
• Decision Tree
• Random Forest
• K-Nearest Neighbors (KNN)
• Gradient Boosting
• Neural Network

Model Performance

The top three machine learning models, ranked by accuracy, are as follows:

1. Gradient Boosting Classifier – Achieved an accuracy of 87%
2. Random Forest Classifier – Achieved an accuracy of 86%
3. Decision Tree Classifier – Achieved an accuracy of 82%

Evaluation

The models were evaluated using several performance metrics:

• Accuracy
• Precision
• Recall
• F1-score

These metrics helped determine the effectiveness of each model in predicting the energy range of solar flares.

Findings

• The Random Forest and Gradient Boosting models achieved the highest accuracy in predicting the energy range of solar flares.
• The project's findings contribute to the field of space weather research by providing a potential tool for predicting solar flares, which is critical for understanding and mitigating their impact on Earth.

Conclusion

This project showcases the application of machine learning to space weather research, with a focus on predicting solar flares. The success of the Gradient Boosting and Random Forest models highlights their potential as robust tools in the field. Through this project, our team, Arjuna, aims to contribute to the advancement of solar flare prediction capabilities.

Acknowledgments

• This project was developed as part of the NASA Space Apps Challenge 2024 by Team Arjuna.
• Special thanks to the NASA RHESSI Data Repository for providing the data used in this analysis.

Team Information

To learn more about Team Arjuna and its members, please visit our team page here.

References

• NASA RHESSI Data Repository
• FITS Viewer and Tools
• Dataset Compilation Script
• Complete Dataset