Study on outliers in the big stellar spectral dataset of the fifth data release (DR5) of the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST)

To study the quality of stellar spectra Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) and correctness corresponding parameters derived by LASP (LAMOST Stellar Parameter Pipeline), outlier analysis method is applied to archived AFGK stars in fifth data release (DR5) LAMOST. The factor defined order sort more than 3 million selected from DR5 catalog. We propose an improved Local Outlier Factor (LOF) based on Principal Component Analysis Monte Carlo enable computation LOF rankings for randomly picked sub-samples that are computed parallel multiple computers, finally obtain ranking each spectrum entire dataset. Totally 3,627 most ranked spectra, around one-thousandth all clustered into 10 groups, parameter density distribution them conforms LAMOST DR5, which suggests whole space probability bad uniformly distributed. By cross-matching with APOGEE, we 122 common ones. published calculated agree APOGEE although there pixels or flux calibrations them. On other hand, some show strong nebular contamination warning should be carefully used. A catalog a spectral atlas outliers can found at link http://paperdata.china-vo.org/LY_paper/dr5Outlier/dr5Outlier_resource.zip.