Accurate determination of the stellar atmospheric physical parameters is crucial for Galactic archaeology via large-scale spectroscopic surveys. Because of the high dimension of celestial spectra, the computational complexity of processing them is relatively high. In this paper, we use the physical signature of each stellar spectrum—line indices as the input features of the Extremely Randomized Trees model (ERT) to estimate the atmospheric physical parameters. We firstly do the correlation analysis between the 26 line indices with the three stellar atmospheric parameters— effective temperature Teff, surface gravity log g and metallicity [Fe/H], respectively. And then, according to the correlation analysis results, we use different combinations of the line indices as the input features of the ERT model to estimate the three parameters. Finally, we analyze the interaction of the peak wavelengths for the three physical parameters, and we found that it could affect the estimation precision of the Teff and [Fe/H], but not the log g. Base on the above the analysis results, we obtained the three parameters with a precision of 59.17 K for Teff, 0.071 dex for log g, 0.0971 dex for [Fe/H]. At the same time, we design a new method to estimate three physical parameters. The results show that this method can be applied to estimate the stellar atmospheric physical parameters effectively.
[1]
Yanchun Liang,et al.
Automatic determination of stellar atmospheric parameters and construction of stellar spectral templates of the Guoshoujing Telescope (LAMOST)
,
2011,
1105.2681.
[2]
Jingchang Pan,et al.
Stellar atmospheric parameter estimation using Gaussian process regression
,
2015
.
[3]
Leo Breiman,et al.
Random Forests
,
2001,
Machine Learning.
[4]
Alexander Szalay,et al.
Photometric Redshift Estimation on SDSS Data Using Random Forests
,
2007
.
[5]
Xiangru Li,et al.
LINEARLY SUPPORTING FEATURE EXTRACTION FOR AUTOMATED ESTIMATION OF STELLAR ATMOSPHERIC PARAMETERS
,
2015,
ArXiv.
[6]
Xiangru Li,et al.
Estimating stellar atmospheric parameters based on LASSO and support-vector regression
,
2015,
1508.00369.
[7]
Pierre Geurts,et al.
Extremely randomized trees
,
2006,
Machine Learning.
[8]
M. Tsantaki,et al.
New Teff and [Fe/H] spectroscopic calibration for FGK dwarfs and GK giants
,
2016,
1608.08392.
[9]
Wei Zhang,et al.
Estimating stellar atmospheric parameters, absolute magnitudes and elemental abundances from the LAMOST spectra with Kernel-based principal component analysis
,
2016,
1610.00083.