Predicting solar flares with machine learning

Large flares and coronal mass ejections (CMEs) often result in dramatic consequences for the functioning of a wide range of ground-based (pipelines, power lines) and space-based infrastructures and services (satellites, GPS), as they may be seriously damaged. Therefore, the importance and needs of flare/CME forecasting is now rapidly increasing and approaches critical levels for the security of our technosphere.
First, I give a brief summary of how we could monitor the occurrence of solar flares and identify features that help predict this phenomenon. Being able to predict them in real-time is important because flares erupting on the Sun impact Earth over the course of minutes.
Nowadays, machine learning techniques provide a rapid, continuously updated overview of the state of the source region of flares, therefore, we also can get alert in real-time about an upcoming large flare/CME event. However, unfortunately, we cannot use machine learning techniques as a magic box in this research field, which we discuss and demonstrate through examples.