Comparing Feature Extraction of Variable Star Light Curves Using Raw and Detrended/Cleaned Data

Variable stars such as RR Lyrae and Cepheids are crucial in astronomy to understanding properties of stars and the evolution of the Milky Way. These stars can be identified manually by their light curves, but the amount of raw data makes this time consuming and ineffective. Using data from sector 1 of the TESS survey, our team aims to test the effectiveness of machine learning in identifying and classifying types of variable stars. In order to use machine learning algorithms, we must first calculate various features of each star’s light curve. Our team's goal was to test the effectiveness of covariant basis vector (CBV) detrending and median cleaning in improving the calculation of light curve features. We begin by calculating the features of each light curve and use corner plots to compare the distribution of features among star types from the raw as well as the detrended and cleaned data. The results show that while some features show significant changes when using cleaned data, most features do not show significant changes indicating that feature calculation of variable stars is not heavily influenced by using cleaned data but could potentially be improved by applying a data celaning process.