Machine-learning analysis of solar flare light-curve morphology and implications for stellar CME prediction

This repository contains the training and testing datasets, extracted features of solar flares, FGKM-type late-type main-sequence star superflares and Sun-like star superflares, as well as event catalogs, the five trained LR sub-models, and the prediction code.

"train_datasets.zip" and "test_datasets.zip". These files contain the training and testing datasets, respectively. The solar flare events were randomly split into training and testing sets with a ratio of 8:2 using five different random seeds, resulting in five independent subsets. The column names are consistent with the feature extraction descriptions in the paper.

"solar_feature.csv", "FGKM_late_main_sequence_stars_feature.csv" and "sun_like_stars_feature.csv". These files contain the extracted features of solar flares, FGKM-type late-type main-sequence star superflares, and Sun-like star superflares, respectively.

"solar_flare_event_catalog.csv" and "stellar_superflare_catalog.csv". These files represent the catalogs of solar flare events and stellar superflare events. "solar_flare_event_catalog.csv"— SDO/EVE-ESP 0.1-7nm broadband solar flare samples containing 1,156 events. Columns: No. — Event serial number; Start_Time(UTC), Peak_Time(UTC), End_Time(UTC) — Event start, peak, and end times in UTC; ED — Equivalent duration (s); Energy — Bolometric energy (erg); Class_Type — Flare class; CME_Association — 1 = eruptive flare (associated with CME), 0 = confined flare (no CME). "stellar_superflare_catalog.csv" — After quality selection, the final Kepler 30-minutes cadence white-light stellar flare sample contains 17,717 events. Columns: No. — Event serial number; Star_Name — Host star identifier; Start_Time, End_Time — Flare start and end times (BJD_TDB − 2,454,833; days); ED — Equivalent duration (s); Energy — Bolometric energy (erg); CME_Association — 1 = with CME, 0 = without CME; Reference — Source of flare.

"logic_models.zip". This file contains the five trained LR sub-models.

"prediction_code.zip". This file contains the prediction code.