A machine learning project leveraging LSTM, GRU, and Bidirectional LSTM + GRU models for accurate prediction of solar flare peak current per second (c/s) and energy magnitude, with comprehensive hyperparameter tuning for optimal performance

The SWAN-SF dataset is now fully preprocessed, optimized, and ready for binary classification tasks. Our team is excited to release the enhanced version of the SWAN-SF dataset across all five partitions.

Accurate solar flare prediction is crucial for mitigating risks to astronauts, space equipment, and satellite communication. Our study enhances prediction accuracy using advanced preprocessing and a novel deep learning-based classifier called ContReg on the SWAN-SF dataset, outperforming previous methods.

These notebooks provide a comprehensive workflow, from start to finish, for processing and analyzing the SWAN-SF dataset. They include detailed steps for reading the dataset files, performing full preprocessing, and executing classification.

This is an NSF-funded project with the goal of developing a machine learning-based web service similar to ChatGPT. This platform will enable users to train AI models for solar flare prediction without requiring any coding expertise.

This repository contains code and data for predicting solar flare energy ranges using machine learning, based on NASA's RHESSI mission data. It includes preprocessing of FITS files into a unified CSV dataset and implements models like Gradient Boosting, Random Forest, and Decision Tree classifiers, achieving accuracies up to 87%.

By analyzing the results of this project, we can identify the most effective data preprocessing techniques and classifiers, ultimately leading to the development of a highly accurate solar flare prediction model.

CONTREX, a novel contrastive representation learning approach for multivariate time series data.