科研之友
微信
新浪微博
Facebook
分享链接
摘要
Variable selection is widely utilized in spectral data analysis. However, for the quantitative analysis of laser-induced breakdown spectroscopy (LIBS), conventional variable selection methods could be inapplicable due to too many dense and informative variables leading to unreliable modeling and a bad explanation effect. To solve this problem, we propose a fast variable selection method combining interval partial least squares (iPLS) and modified iterative predictor weighting-partial least squares (mIPW-PLS) and apply it to the LIBS quantitative analysis of soils. Our method is validated by detecting the concentrations of Cu, Ba, Cr, Mg and Ca in different kinds of soil samples. The results demonstrate that variables with maximal relevance were selected effectively and efficiently. Compared with other methods, our method establishes a more simplified model with high accuracy and robustness as well as improved predictive ability. The number of variables used for calculation was reduced significantly. The root mean square error of prediction (RMSEP) and the R-2 of prediction showed more satisfactory results compared with conventional methods. The limits of detection (LODs) obtained for Cu, Ba, Cr, Mg and Ca were 11.4, 4.3, 3.6, 529.5 and 307.6 mg kg(-1).
