Premium
Mixture analysis using non‐negative elastic net for Raman spectroscopy
Author(s) -
Zeng HuiTao,
Hou MingHui,
Ni YiPing,
Fang Zhi,
Fan XiaQiong,
Lu HongMei,
Zhang ZhiMin
Publication year - 2020
Publication title -
journal of chemometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.47
H-Index - 92
eISSN - 1099-128X
pISSN - 0886-9383
DOI - 10.1002/cem.3293
Subject(s) - raman spectroscopy , lasso (programming language) , fingerprint (computing) , biological system , stability (learning theory) , elastic net regularization , analytical chemistry (journal) , relative standard deviation , chemometrics , chemistry , chromatography , materials science , pattern recognition (psychology) , artificial intelligence , computer science , physics , machine learning , optics , feature selection , detection limit , world wide web , biology
Raman spectrum is wealth of structural information, which can be used as molecular fingerprint to identify compounds. There are relations between the intensities of Raman peaks and the concentrations of compound, and they can be used to estimate relative concentrations of the compounds in mixture. However, it is still challenging to interpret the Raman spectrum of mixture. The non‐negative lasso (NN‐LASSO) has been used to solve this problem, but it failed to identify highly correlated compounds. The quadratic term of non‐negative elastic net (NN‐EN) can ensure the stability of the fitted model. Therefore, a novel mixture analysis method was developed based on NN‐EN in this study. It has been applied to analyze the simulated, liquid, powder, tablet, and quantitative mixture datasets. Results showed that NN‐EN can identify the compounds in mixture with high accuracy and estimate their relative concentrations with small deviation. Furthermore, NN‐EN was more stable than NN‐LASSO when the spectra of some compounds are highly correlated. It is a promising approach for analyzing Raman spectra of mixtures.