Regression Prediction of Effective Temperature of Stellar Spectrum Based on Kernel Method

Abstract

There are many factors that lead to the huge difference in stellar spectra, the most important of which are atmospheric physical parameters, namely, effective temperature, surface gravity and chemical abundance. This study is based on the classified star data in LAMOST DR6 with a temperature value of 7500-9000K and a signal-to-noise ratio of S/N greater than 50. The results of the two methods are compared through the regression verification of nuclear least squares regression (KLSR) and nuclear PCA regression (KPCR), The scope of application of the two methods is discussed, and the two most important lick parameters affecting the effective temperature are regressed. By regressing the effective temperature of such stars, a method for predicting the effective temperature of measured spectral data under large samples is given, which provides a certain reference for predicting the trend of star evolution and studying the evolution law of such stars.