© 2020 All Rights Reserved

Component analysis and informative feature selection for Raman spectral data

Y.V. Kistenev1, D. A. Vrazhnov1,2*, O. P. Cherkasova3,4*, A. A. Mankova5, Karmenyan A.V.6, Perevedentseva E.V.6,7, Krivokharchenko A.S.8, Sarmiento M.N.6, Barus E.L.6, Cheng C.-L.6, Malakhova T.9, Kabanova T.9
1Tomsk State University, Russia
2Institute of Strength Physics and Materials Science of Siberian Branch of the RAS, Russia
3Institute of Laser Physics SB RAS, Russia
4Institute on Laser and Information Technologies - Branch of the Federal Scientific Research Centre "Crystallography and Photonics" of RAS, Russia
5Lomonosov Moscow State University, Russia
6Department of Physics, National Dong Hwa University, Hualien, Taiwan
7P.N. Lebedev Physics Institute of Russian Academy of Sciences, Moscow, Russia
8N. N. Semenov Institute of the Chemical Physics, RAS, Moscow, Russia
9 Institute of Applied Mathematics and Computer Science, Tomsk State University, Tomsk, Russia


The application of machine learning methods to Raman spectral data analysis allows constructing new data models, but the relations between input features and output remain hidden. Informative feature selection (IFS) methods are used to find latent relations in complex data models. In this work, we investigate different types of IFS, such as wrapper-based and embedded methods, and illustrate their advantages and limitations in application to experimental Raman spectral data.

This work has been supported by Russian Foundation for Basic Research (project № 17-00-00275 (17-00-00270, 17-00-00186)). The work was performed under the government statement of work for ISPMS Project No. III.23.2.10. AVK, EVP, CHLCH, EB, MS were supported by grants from the Ministry of Science and Technology of Taiwan, MOST 108-2923-M-259-002 -MY3 and ASK, RFBR grant 19-53-52007

File with abstract


Denis Vrazhnov
Institute of Strength Physics and Materials Science of Siberian Branch of Russian Academy of Sciences


File with report


Ask question