Predicting Solar Flares Using CNN and LSTM on Two Solar Cycles of Active Region Data

This repository contains codes for data processing, model training, and performance evaluation used in paper Predicting Solar Flares Using CNN and LSTM on Two Solar Cycles of Active Region Data (Sun et al. 2022)

Quick start

1. Download data: Change the email and data directory in download.py and run python download.py.
2. Preprocess data
3. Change the data directory in preprocess.py.
4. Install Redis. (Alternatively, change the default value of redis to False in function query in data.py)
5. Run python preprocess.py.
6. Exploratory data analysis (eda.py)
7. Fit and evaluate machine learning methods:
8. Scikit-learn models
9. Pytorch-lightning models 1. MLP 2. LSTM 3. 2D CNN 4. 3D CNN
10. Present results (notebooks/mlflow_results.ipynb)

Citation

@article{sun2022predicting,
  title={Predicting solar flares using cnn and lstm on two solar cycles of active region data},
  author={Sun, Zeyu and Bobra, Monica G and Wang, Xiantong and Wang, Yu and Sun, Hu and Gombosi, Tamas and Chen, Yang and Hero, Alfred},
  journal={The Astrophysical Journal},
  volume={931},
  number={2},
  pages={163},
  year={2022},
  publisher={IOP Publishing}
}