A calibration set selection method fusing multi-component based on Euclidean distance in spectral analysis of complex solution

Abstract

In the spectrochemical quantitative analysis of complex solutions, selecting the appropriate calibration set selection method is important for the model's universality. Based on Lambert-Beer law and “M + N” theory, this paper proposes a calibration set selection method fusing multi-component based on Euclidean distance. This method comprehensively considers the contribution of each component to the spectrum, and uses the weighted Euclidean distance as the index to select the calibration set. To prove this method, the sample set partitioning strategy based on multi-component spatial distance (SSP-MCSD) and the calibration set selection method considering multi-component fusion are used to model the six primary components of 404 samples respectively in this paper. The results show that the method in this paper yields better modeling results than the original calibration set selection method. Among them, the related coefficient of calibration set (Rc) of hemoglobin modeling was increased from 0.771098 to 0.825107, and the difference between the Rc and the related coefficient of prediction set (Rp) was reduced from 0.048943 to 0.015074. The Rc of red blood cell modeling was increased from 0.741572 to 0.808852, and the difference between the Rc and the Rp was reduced from 0.088833 to 0.015306. The modeling results for other components were also improved, for example, the Rc of white blood cells modeling was increased from 0.189402 to 0.231586, which was improved by 22.2%. In general, the method proposed in this paper fully considers the distribution of non-targeted components in the calibration set, and the accuracy and robustness of the model are better than the original calibration set selection methods. It also provides a new idea for the spectrochemical quantitative analysis of complex solutions.