Using Unsupervised Machine Learning for Distinguishing Stellar Explosions from Eruptions

Abstract

Our goal for this paper is to investigate the relationships between the properties of supernovae within the ZTF BTS sample. We applied unsupervised clustering methods to time-domain astronomical data and to distinguish supernovae and stellar eruptions. Testing different models, we discovered that K-Means provided the best results, clearly distinguishing 2 clusters compared to the rest. While we used a small sample of data for the model, this could be applied to larger pieces of data as well. Since this tool is very effective in distinguishing different types of astrophysical phenomena, this research could also enhance astrophysics knowledge in the lengths of the stellar and the elemental definition of stars at specified evolutionary stages.