Miniaturization of analytical technology has paved the way for in-situ screening of foods. In the current study, the spectral features of olive oils are examined by handheld near-infrared spectroscopy to explore the technology's capabilities to distinguish extra virgin olive oil (EVOO) from lower grade oils. Eighty EVOO, forty refined olive oil (ROO), and ten pomace olive oil (POO) samples are analysed for their spectral and compositional features. The latter included analysis of the fatty acids (FAs), the chlorophylls and carotenoids, chromatic coordinates and moisture contents. The 1350–1570 nm wavelength range appeared most suitable for distinction of the oils. One-class classification models with three different classifiers are subsequently estimated using this range, and their quantitative performance is assessed from probabilistic data. Soft independent modeling of class analogies models appears to predict the identity of the oils with a high success rate. Compared to the other oils, POO comprises a significantly higher and lower proportion of polyunsaturated and monounsaturated FAs, respectively. Higher contents of chlorophylls, carotenoids, and moisture are noted for EVOO. The relevant spectral information for distinction of the oils correlates strongly with the degree of unsaturation of the oils as well as their levels of chlorophylls, carotenoids, and moisture. Practical Applications: The findings of this study demonstrate that the handheld NIRS technique is promising for future rapid screening of olive oil grades. The statistical methods used and the robust validation procedure will help potential users to select the optimal strategy for multivariate data analysis. In addition, the exploration of correlations with compositional characteristics provides insight into the handheld NIRS working mechanism in regard to EVOO authentication.
- food fraud
- one-class classification