Using Machine Learning for Automatic Classification of Classical Cepheids


With the increasing amounts of astronomical data being gathered, it is becoming more crucial for machine learning techniques to be employed for star classification. Classical Cepheid variable stars can be grouped into several classes, such as fundamental-mode, first-overtone, and second-overtone. Each class has distinctive features, and the light curves of the stars can be analyzed for these features in order to be used in automatic classification. Here, we focus on developing a number of features to be used with the following machine learning methods: Multilayer Perceptron, Naïve Bayes, J48 Decision Trees, and Random Forest. We use the OGLE (Optical Gravitational Lensing Experiment) datasets of Classical Cepheid variable stars in the Large Magellanic Cloud and the Small Magellanic Cloud. Our findings indicate that the Multilayer Perceptron is an excellent method for approaching this problem, as it outperformed the other machine learning methods. We also identify a number of useful features using Information Gain and Gain Ratio. Specifically, the newly developed features to measure symmetry had high classification power.



Machine learning, Classification, Large Magellanic Cloud, Small Magellanic Cloud, Cepheids, Cepheids, Variable stars, Variable stars, Astronomy, Computer science, OGLE, Data, Fundamental mode, First overtone, Second overtone, Fundamental mode, First overtone, Second overtone, Multilayer Perceptron, Neural networks, Neurosciences, Information gain, Gain Ratio, Classifiers, Star, Stars