Solar flares are powerful bursts of energy from the sun that can disrupt communication systems, GPS, and even power grids on Earth. Predicting when these flares will happen is important for preparing for and minimizing their impact. Scientists use computer models to help make these predictions, but fine-tuning these models to make them accurate is often slow, complex, and uses a lot of energy.

In this study, we explore a new way to make the prediction process faster and more efficient by using visual tools to guide human decision-making. Instead of relying entirely on computers to search through countless combinations of model settings, we allow scientists to visually explore the model’s performance and focus only on the most promising areas. This human-in-the-loop approach can save time, reduce energy use, and still produce accurate results.

We tested this method using solar data collected in the hours leading up to flares, trying to predict flare events up to 24 hours in advance. Our visual approach showed that scientists could quickly identify good settings for the model. This technique could also be helpful for other types of time-based predictions beyond solar flares.