Rapid Determination of Peroxide Value of Peanut Oils During Storage Based on Terahertz Spectroscopy

Abstract

Terahertz (THz) spectroscopy was used to evaluate peroxide value (PV) of peanut oils during storage. Genetic algorithm (GA) and principal component analysis (PCA) were used as the pretreatment methods to reduce THz high-dimensional data. Interval partial least square (iPLS), back propagation neural network (BPNN), support vector regression (SVR), and partial least squares regression (PLSR) were compared to obtain the best quantitative prediction model. The optimal absorbance spectral region related to peanut oils oxidation was found from 1.1749 to 1.5030 THz by iPLS method. In all chemometric methods used, PCA-SVR showed the best quantitative correlation between absorbance spectra and PV, with the correlation coefficient in prediction set (RP) was 0.9289 and root-mean-square error (RMSEP) was 1.6648 meq O2/kg, respectively. It can be concluded that THz spectroscopy with chemometric methods was an effective tool for rapid determination of PV of peanut oils.