Grid clustering analysis the big data of spectrum by deep learning

Abstract

With the Internet plus big data science information era increasing rapidly. To seek special and unknown objects is the human exploration of the mystery of the universe to pursue the goal in the universe. The spectrum by the big data mining are the fairly complex data, the dimension is high, and the correlation between the dimensions is not strong, but it is easy to introduce noise or the missing data. So it is much more difficult to deal with metering data. This article investigates the LAMOST data release star spectrum based on the high resolution spectral parameters. The RFITSIO software package of R language is used to graphically analyze the big data of the spectrum. The deep learning analysis extracts the information from the large data with finding the new knowledge and the unknown outlier data. Now the FITS format spectral large data information rise to 107 levels of data. Since the big data is imported with a large amount of redundant information, the full spectrum signal of the star spectrum making the full use of Multivariable Statistical Analysis to cluster clustering data characterized by line index. Using the Lick line index as the spectral feature, the spectral data are clustered by the K-means mean algorithm of deep learning. Experiments show that the data with strong physical correlation are valid and fast, the clustering outlier analysis of the big data feature in the spectral survey are completed with the characteristics of the data.