Classification of Stellar Using Supervised Machine Learning Techniques

In this paper, we show that machine learning methods may be successfully used to solve the fundamental stellar astronomy challenge of predicting star atmosphere properties from spectral data. In order to enable their decision-making, stellar-atmospheric refraction (SAR) adjustments must be quick, autonomous, and precise for star trackers working in the atmosphere. The article's presentation of a combined approximation both SAR & stargazer attitude (often referred to as S) satisfies this need. A SAR model has been developed using the corresponding refractive coefficients (ERC) and true zenith distances (TZDs). It is recommended to do combination SAR and STA estimate utilizing frequent errors replication (REF) cycle. In contrast to current rectification methods, this proposal avoids the problems involved with theoretical environmental simulations and related empirical models. We use Kohonen identity maps, forward-feeding neural networks, and locally weighted regression to derive actual temperatures, surfaces gravity, or metallic stellar atmospheric properties from spectral indices. Our test findings show that each of the three techniques is very good at predicting parameters. Weighted basis regression outperforms other ways that use the original information as input by a small amount, whereas all self-planning schemes perform better than other techniques while generating a lot more noise. To combine the results from our algorithms, we also use a variety of predictors. When using unprocessed and noisy data, the cluster outperformed every one of the three approaches.