Reducing collinearity by reforming spectral lines with two-dimensional variable selection method

In the spectral quantitative analysis of complex solutions, the broad width of the spectral peak comprises the multi-component spectra, which mix up with neighboring characteristic bands and cause collinear-ity of spectra and seriously affecting the accuracy of the analysis. A two-dimensional variable selection method is proposed to solve this problem by reforming the component spectrum in this paper. The use-ful data is selected from the spectral dataset with two dimensions of wavelength and pathlength which together represent the chemical and physical properties of the component. The new spectra of multi-ple components are formed with these data, and they are different from each other to eliminate the collinearity. The method's effectiveness is studied through the quantitative analysis of four components in human blood. According to one-dimensional and two-dimensional variable selection methods, interval partial least squares (iPLS), genetic algorithm PL S (GA-PL S), and ant colony optimization PL S (ACO-PL S) were used to select the effective datasets of four components, respectively. The experimental results show that the model's prediction accuracy based on the two-dimensional variable selection method is much better than that based on one-dimensional variables.(c) 2022 Published by Elsevier B.V.

Automated summary

This paper proposes a two-dimensional variable selection method to solve the problem of spectral collinearity in complex solutions by reforming the component spectrum. The effectiveness of the method is studied through quantitative analysis of four components in human blood, and the results show that the prediction accuracy based on the two-dimensional variable selection method is superior to that based on one-dimensional variables.