Determination of the Oxidative Stability of Camellia Oils Using a Chemometrics Tool Based on 1H NMR Spectra and α-Tocopherol Content.

This study, for the first time, predicts oxidative stability in camellia oils by partial least squares (PLS) built with proton nuclear magnetic resonance (1H NMR) and α-tocopherol content. The prediction models were established by the PLS method. Outlier detection, latent variables optimization, data pretreatment, and important variables selection were applied for models optimization. All the developed models exhibited good performance as indicated by R2 > 0.895 and root mean square error of estimation and root mean square error of prediction less than 0.322 and 0.307. For verification of the contribution of 1H NMR spectra and α-tocopherol for prediction performance, a PLS model with fatty acids composition instead of 1H NMR spectra and one with only 1H NMR spectra as input variables were developed, respectively. The results showed that the model based on 1H NMR data was more accurate and precise than that based on fatty acid composition data. And the performance of the models was significantly degraded without α-tocopherol as input variables.