Stellar Objects Classification Using Supervised Machine Learning Techniques

Abstract

Machine Learning is used in many fields of study. This paper used machine learning to classify instances from the Sloan Digital Sky Survey Data Release 17 (SDSS DR17) as a galaxy, quasar, or star. Supervised learning was used to make the classification. Multiple machine learning models were built, Decision Trees, K-Nearest Neighbors, Multinomial Logistic Classification, Multilayer Perceptron, Naïve Bayes Classifier, Support Vector Classification, Random Forest, and Soft Voting Classifier. Random Forest performed the best with 98% accuracy and correctly classified all instances labeled as stars in the dataset. The worst-performing algorithm was Naïve Bayes, with 91% accuracy.