Identifying Variable Stars from Kepler Data Using Machine Learning

Abstract

Machine learning algorithms play an impressive role in modern technology and address automation problems in many fields as these techniques can be used to identify features with high sensitivity, which humans or other programming techniques aren’t capable of detecting. In addition, the growth of the availability of the data demands the need of faster, accurate, and more reliable automating methods of extracting information, reforming, and preprocessing, and analyzing them in the world of science. The development of machine learning techniques to automate complex manual programs is a time relevant research in astrophysics as it’s a field where, experts are dealing with large sets of data every day. In this study, an automated classification was built for 6 types of star classes Beta Cephei, Delta Scuti, Gamma Doradus, Red Giants, RR Lyrae and RV Tarui with widely varying properties, features extracted from training dataset of stellar light curves obtained from Kepler mission. The Random Forest classification model was used as the Machine Learning model and both periodic and non-periodic features extracted from light curves were used as the inputs to the model. Our implementation achieved an accuracy of 86.5%, an average precision level of 0.86, an average recall value of 0.87, and average F1-Score of 0.86 for the testing dataset obtained from the Kepler mission.

Feature-Based Classification Neural Network for Kepler Light Curves from Quarter 1

“…In the literature, the classification of light curves usually relies on the periodic features from the light curves [8,15], which has two main challenges. Firstly, temporal gaps and noise in the light curves can impact the accurate determination of periods.…”

“…For traditional pattern recognition methods, complex calculations are required to reduce the dimensionality of the light curves, such as the Fourier transform [2][3][4], Kepler photometry features [5], amplitude and coherence time scale features [6], or using the Markov Chain Monte Carlo Technique [7]. Machine learning methods are currently popular in astronomy, which includes supervised learning [8][9][10][11][12] and unsupervised learning [13,14].…”