With the vigorous development of the big data astronomical calculation science, our country has obtained independent intellectual property rights of the international astronomical community.LAMOST telescope has taken the lead in the world in terms of development and implementation of combining the observation with mass survey work of thousands of celestial spectra. With the largest caliber and spectrum which has the highest rate of wide field. Druing the period between 2011 and June 2015, DR3 survey spectra obtained data set to acquire the biggest in the world currently parameters star catalog. In allusion to LAMOST the third of FGK stellar spectra data released the flow-calibration, etc. Kurucz template spectrum was used to partition the mesh, based on open source efficient R language programming supervised data processing software platform , in order that clustering center is easy to verify the theoretical parameters of the grid difference. After normalized processing of LAMOST measured spectrum, on the theoretical parameters grid with the supervised clustering validation, Euclidean distance discriminant algorithm was adopted to distinguish the similarity between the spectra. Choosing distance quantity directly described the attributes, properties of vector constructor judged the observed difference spectroscopy and the theoretical levels were selected. Experiments show that the measured spectrum of LAMOST in FGK type stellar spectra data comparisons with Kurucz theoretical template library data, which has a better consistency and accuracy in parameter measurement with the same parameters between the spectral characteristic lines. Thus, it is concluded that the spectrum measurement physical parameters of LAMOST is higher than the observation quality with excellent reliability, which provides the basis argument for improving the following stellar atmosphere models.