/Solar-Flare-Prediction-through-Time-Series-Data-Augmentation

Solar Flare Prediction through Time Series Data Augmentation

Primary LanguagePythonMIT LicenseMIT

Data-augmentation-for-solar-flare-prediction

This is the repository for our work titled "Solar Flare Prediction through Time Series Data Augmentation", which has been submitted to AGU Space Weather Journal.

Data Download

download the raw data from: https://dmlab.cs.gsu.edu/solar/data/data-comp-2020/

About

This project includes four parts: 1. Data preprocessing (Data preparation) 2. Using different data augmentation methods to generate synthetic samples 3. Apply 3 deep learning models to evaluate the solar flare prediction in different scenarios: (1) undersampling and imbalanced data comparison (2) data augmentation and imbalanced data comparison (3) synthetic/real ratios vs performances (4) Binary classification between synthetic and real 4. Case study on a multi-class classification