Solar Flare Prediction Dataset: SHARP Magnetic Field Parameters for Machine Learning

This dataset contains processed SHARP (Solar Dynamics Observatory/Helioseismic and Magnetic Imager) magnetic field parameters for solar flare prediction tasks, covering the period from January 1, 2023, to May 31, 2025. The dataset follows the data assembly protocol described in Abduallah et al. (2023) Scientific Reports for operational solar flare prediction.

The dataset includes 18 CSV files with training and testing data for three prediction tasks (C-flare, M-flare, M5-flare) across three temporal windows (24h, 48h, 72h), along with normalization parameters and reproduction scripts. Each sample contains nine magnetic field parameters from SHARP data: USFLUX, TOTUSJH, TOTUSJZ, MEANALP, R_VALUE, TOTPOT, SAVNCPP, AREA_ACR, and ABSNJZH.

The dataset uses a chronological 90%/10% train/test split and includes ~273,000 samples total across all tasks. Quality control filters ensure Central Meridian Distance ≤ 70°, radial spacecraft velocity ≤ 3500 m/s, and no missing values in magnetic field parameters.