Development of a pipeline for the analysis of light curves with machine learning techniques

Abstract

Machine learning techniques have been more and more used in astronomy, due to the volume and complexity of observational data taken with different instruments. A possible application of such techniques involves analyses of the so-called light curves, which are data sets used to monitor the variability of the brightness of stars. The analysis of light curves allows to detect possible eclipses, which occur when an object, orbiting the observed star, passes between the star and the observer. Such orbiting body can be an exoplanet or another star, which is called eclipsing binary. In this work, we will evaluate, using data taken with Kepler space telescope, the most efficient machine learning techniques to classify light curves in: curves with eclipses caused by exoplanets, curves with eclipses caused by eclipsing binaries, and curves without eclipses. Based on that and on the results of previous Scientific Initiation projects, we will elaborate a pipeline for the analysis of light curves taken with different instruments. This pipeline will include the use of filtering techniques, to remove high-frequency noise, and the classification of light curves, using machine learning techniques.