Time Series Data of SHARP Magnetic Characteristic Parameters at Neutral Lines of Active Regions for Solar Flare Forecasting

The original data are sourced from the SDO data product called Space-weather HMI Active Region Patches (SHARP), which automatically identifies AR patches from the vector magnetic field data. The total amount of data used in this study exceeds one million records. The data covers 1,290 solar active regions (HARPNUM ranging from 1 to 7,325; IDs assigned chronologically) and includes all available HMI SHARP data from this period, spanning approximately one solar cycle from May 1, 2010, to November 29, 2021 (roughly solar cycle 24). In the flare forecasting task, considering the scarcity of data available for training, we set the test-to-training data ratio to 19 to allow the model to learn from more cases. About 130 active regions serve as the test set, while 1,160 are in the training set.We divided the 1,290 active regions numbered from 1 to 7325 into 10 groups, each containing 129 active region numbers. Each group is further split into 10 partitions, called "folds" (each fold containing 12 or 13 active region numbers) for cross-validation. For the experiment with the N-th fold, the test set consists of the N-th fold from each group, while the remaining active regions comprise the training set.