Quantitative analysis of Cr in soil based on variable selection coupled with multivariate regression using laser-induced breakdown spectroscopy

Abstract

Laser-induced breakdown spectroscopy (LIBS) can be used to quantitively analyze heavy metal pollutants in soil if the interaction of Fe and other matrix elements and the self-absorption effect are eliminated. In spectral data processing, multivariate regression can reduce the interaction between elements and improve the predictive performance of the quantitative model, but still needs to control the overfitting and limit the number of variables of the full spectra. In this work, the adaptive least absolute shrinkage and selection operator (ALASSO) is adopted to select the variables in the sample soil spectra and combine with the support vector regression (SVR) for the quantitative analysis of the Cr content. Compared with the combination of the least absolute shrinkage and selection operator (LASSO) and SVR, the correlation coefficient of the combined model of ALASSO and SVR increases from 0.987 to 0.998. The root mean square error of the calibration set (RMSEC) and the root mean square error of the verification set (RMSEV) are reduced from 0.043 wt% and 0.039 wt% to 0.017 wt% and 0.033 wt%, respectively. Meanwhile, the relative standard deviation (RSD) of the model is reduced from 3.436% to 2.343%. The results show that the combination of ALASSO and SVR improves the accuracy and precision of the quantitative analysis of Cr in soil.