Data mean and ratio of absorbance to concentration methods: A novel optimization strategy for near infrared...

For the modeling of near infrared spectroscopy (NIRS), the accuracy of the basic data, the stability of the spectra and the optimality of variables selection method were the important factors. In this paper, a novel optimization strategy for NIRS modeling was proposed, which was formed by data mean and ratio of absorbance to concentration (RATC) methods. The data mean method was aim to obtain accurate basic data and stable spectra, the RATC method was aim to select the optimal variables and compared with other variables selection methods (FiPLS, BiPLS, CC, UVE). The experimental subject was raw human plasma, with this novel optimization strategy, the predictive capability of NIRS model of its total protein (TP) content had been improved. At the same time, the public NIRS testing data (water, protein, oil, starch of corn and octane of gasoline) were used to verify the proposed variables selection method, and the predictive capability of these models of different parameters were also improved. To some extent, the optimization strategy of NIRS modeling provided theoretical supports for the development of protein content analyzer of NIRS and the quick determination of parameters of biologics and other materials.