Machine Learning Applied to Star–Galaxy–QSO Classification and Data Visualization

With the continuous development of modern astronomical observation methods, the sky survey data obtained through observation has increased exponentially, and machine learning has gradually replaced traditional scientific computing methods with its powerful computing power in the big data environment. In view of the problems of low classification accuracy, poor classification efficiency and few classification types in the current astronomical spectrum classification work, this study added sub-classification to the original classification benchmark and compared the performance of different machine learning algorithms in the classification problem in detail. Visual presentation of data improves the intuitiveness of data. In this study, one hundred thousand LAMOST DR6 spectral data were used for training, and different machine learning algorithms were used to evaluate the performance of various classifiers, and then correlation analysis was used to greatly improve the classification accuracy of specific categories on the basis of the original research, making the classification accuracy of galaxies and QSO is higher than 98%, which proves the validity of this research. Finally, various attributes of the classified celestial objects are displayed through visualization methods.