Classification of Quasars, Galaxies, and Stars by Using XGBoost in SDSS-DR16

Quasar, galaxies, and stars identification is significant and fundamental for astronomy study, it is compelling enough to process these escalating data. The main objective is implementing a highly time-efficient machine-learning algorithm to perform quasars, galaxies, stars classification in the Sloan Digital Sky Survey Data Release 16 (SDSS DR16) spectra dataset. The study uses a gradient boosting machine learning algorithm called XGBoost (Extreme Gradient Boosting) [1]. Implementing this machine learning model requires data selecting, data analysis, data splitting, features extraction, features filtering, feature scaling, hyperparameter tuning, training via XGBoost, and finally getting the accuracy score to evaluate the classifier. Tuning hyperparameters by trial and error is extraordinarily time-consuming, so the author exploited automatic tunning technology (Bayes searching and Bayesian optimization).This paper roughly introduced each algorithm implemented in the test. The author got an accuracy score of 99.39 % in the test set at last.