An iterative algorithm for determining the temporal characteristics of solar X-ray flares for the automated formation of a data set in machine learning systems

Modern astronomical instruments and devices in space provide scientists with a wealth of scientific and technical information. Scientists and developers are confronted with the question of how to gather, process, and store data that defines a particular phenomenon while developing artificial intelligence systems. An extensive array of data processing techniques is also employed in X-ray astronomy; however, the efficacy of these techniques is not universally reliable, which has consequential implications for the overall functionality of artificial intelligence systems and the likelihood of accurate classification. Automation and manual processes constitute the two broad categories into which these methods can be categorized. Each of these categories possesses distinct merits and demerits. Simplyified metric selection and precise definition are factors that contribute to such errors. This study examines the process of determining temporal metrics for the automated creation of a data set using the light curves of solar X-ray flares. The acquired data is designed for eventual utilization in machine learning systems. In order to determine the temporal characteristics of a solar X-ray flare, the flare's initiation, maximum, and termination points are postulated. The provided quantity of data points is adequate for ascertaining the X-ray burst's total duration, rise time, and decay time. Using an iterative algorithm, the authors suggest making it possible to automatically figure out the metrics related to the start, peak, and end of an X-ray flare. Researchers are testing the iterative algorithm on fake data made by the damped oscillations function, on fake periodograms that make the light curve more accurate, and on real data from solar X-ray bursts. This approach enables the automated extraction of temporal attributes of a solar X-ray burst for subsequent storage and aggregation as a database. It also facilitates further processing and utilization in the training of artificial intelligence systems.

This work is supported by the “long-term program of support of the Ukrainian research teams at the Polish Academy of Sciences carried out in collaboration with the U.S. National Academy of Sciences with the financial support of external partners.”