| Abstract * |
: |
With machine learning entering into the awareness of the heliophysics community, solar flare prediction has become a topic of increased interest. Although the machine learning models have changed with each successive publication, most of the input data is based on magnetic features. Despite increased model complexity results seem to indicate that photospheric magnetic field data alone may not be a wholly sufficient source of data for flare prediction. For the first time we have extended the study of flare prediction to spectral data. In this work, we use Deep Neural Networks to monitor the changes of several features derived from the strong resonant Mg II h&k lines observed by IRIS. The features in descending order of predictive capability are: line Intensity, the logarithm of the ratio of the red wing subordinate line at 2798.77 Å with respect to the continuum at 2799.32 Å, total continuum emission between the h&k line cores, the k/h ratio, line-width, followed by several other line features such as asymmetry and line center. Regions that are about to flare generate profiles which are distinguishable from non-flaring active region profiles. Our algorithm can correctly identify pre-flare profiles 40 minutes before the start of the flare with an 80 % accuracy, precision and recall. This accuracy monotonically increases to 88 % as we move closer in time to the start of the flare. Our study indicates that sequence modeling of spectral data can in itself lead to good predictive models and should be considered as an additional source of information along side photospheric magnetograms. |
