Table 1.
Description of twelve flare forecasting studies based on machine learning.
Paper | Data | Multiple test | AR data split | Validation | Method | Score – C1+ | Score – M1+ |
---|---|---|---|---|---|---|---|
realizations | |||||||
Bobra & Couvidat (2015) | Point in time | Yes | Yes | Yes | SVM | – | 0.74 |
features (SHARP) | |||||||
Liu et al. (2017) | Point in time | Yes | Yes | Yes | RF | – | 0.76 |
features (SHARP) | |||||||
Nishizuka et al. (2018) | Point in time | No | No | No | MLP | 0.63 | 0.80 |
features (ad hoc computed) | |||||||
Florios et al. (2018) | Point in time | Yes | Yes | Yes | RF | 0.60 | 0.74 |
features (FLARECAST) | |||||||
Jonas et al. (2018) | Time series | Yes | No | Yes | RF | – | 0.74−0.81 |
features | |||||||
Campi et al. (2019) | Point in time | Yes | No | Yes | Hybrid lasso | 0.54 | 0.67 |
features (FLARECAST) | |||||||
Liu et al. (2019) | Time series | Yes | No | Yes | LSTM | 0.61 | 0.79 |
features (SHARP) | |||||||
Wang et al. (2020) | Time series | ... | No | yes | LSTM | 0.55 | 0.68 |
features (SHARP) | |||||||
Park et al. (2018) | HMI and | No | No | Yes | CNN | 0.63 | ... |
MDI magnetograms | |||||||
Huang et al. (2018) | HMI and | Yes | Yes | – | CNN | 0.49 | 0.66 |
MDI magnetograms | |||||||
Li et al. (2020) | HMI | Yes | No | No | CNN | 0.68 | 0.75 |
magnetograms | |||||||
Yi et al. (2021) | HMI | No | No | Yes | CNN | 0.65 | – |
magnetograms |
Notes. For each study, the table reports: the main author (column “paper”); the kind of data used (column “data”); whether a confidence strip has been computed for the skill score (column “multiple test realizations”); whether data belonging to the same AR are split between the training and test sets (column “AR data split”); whether a validation set has been used to optimize the machine learning algorithm (column “validation”); which method has been used (column “method”); and the score values for the prediction of C1+ and M1+ flares (columns “score – C1+” and “score – M1+”).
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.