Precursor identification for strong flares based on anomaly detection algorithm

Abstract

In this study, we assume that the magnetic configuration of active regions (ARs) in quiet periods has certain similarities and can be considered “normal” features. While there are some other magnetic features of active regions that are related to strong flares, they can be considered the precursor of strong flares and “anomaly” features. Our study aims to identify those “anomalies” and apply them in strong-flare forecasting. An unsupervised auto-encoder network has been used to understand and memorize these “normal” features, and then, based on the mean squared errors between the pictures of the ARs and the corresponding reconstructed pictures derived by the network, an anomaly detection algorithm has been adopted to identify the precursor for strong flares and develop a strong-flare classification model. The strong-flare classification model reaches an F1 score of 0.8139, an accuracy of 0.8954, a recall of 0.8785, and a precision of 0.7581. Moreover, for those correctly predicted strong-flare events (94 M-class flares and above), the model reaches an average first warning time of 45.24 h. The results indicate that the anomaly detection algorithm can be used in precursor identification for strong flares and help in both improving strong-flare prediction accuracy and enlarging the time in advance. Also, the obtained average maximum warning period for strong-flare prediction (nearly 2 days) will be useful for future applications for space-weather solar flare prediction.