Classification of Long Gamma-Ray Transients from INTEGRAL Data Using Machine Learning Approach

In this paper we use 19 years of data from INTEGRAL detectors to train classification model for gamma-ray bursts. We present algorithms for automated processing of the light curve of gamma-ray bursts, consisting of a background approximation, distinguishing an event from background, and duration calculation. Candidates are crossmatched with several catalogues of transient events. This provided us labels for supervised machine learning. Gradient boosting classifier is employed for training to find Solar flares and gamma-ray bursts. Estimated accuracy is \(\sim \) 91% for latter events. Similar machine learning approach can be applied to other types of transients.