Rapid Determination of Unsaturated Fatty Acids in Vegetable Oil by Raman Spectroscopy and Chemometrics

The accuracy for the determination of total monounsaturated and polyunsaturated fatty acids in vegetable oil by Raman spectroscopy was enhanced by spectral wavelength selection with competitive adaptive reweighted sampling to obtain partial least squares models based on a small subset of wavelengths. Wavelengths of 26 and 31 were selected from the 842 original points for the total monounsaturated and polyunsaturated fatty acids, respectively. Compared to full spectral models and four frequently-used variable selection methods, the competitive adaptive reweighted sampling achieved better results. The external correlation coefficients for total monounsaturated fatty acids and polyunsaturated fatty acids were increased to 0.9555 and 0.9963, and root mean square errors of prediction were reduced to 0.0207 and 0.0054. The results indicate that a portable Raman spectrometer using competitive adaptive reweighted sampling with partial least squares may be a promising approach for the determination of unsaturated fatty acids in vegetable oil.