ABSTRACT
The potential of front-face fluorescence spectroscopy coupled with chemometric techniques, namely multiple linear regression (MLR) applied on parallel factor (PARAFAC) scores and partial least squares (PLS), was tested on Lebanese olive oil samples possessing natural variability within their chemical parameters. Ninety-six olive oil samples have been harvested at different dates and from two seasons, processed using different extraction methods, collected from different altitudes and other factors that can increase the variability of the samples' chemical composition. Fluorescence excitation-emission matrices (EEM) of the collected samples were measured, and the relationship between them and the chemical parameters was examined. Twenty-two MLR regression models based on PARAFAC scores were generated, the majority of which showed a good correlation coefficient (R > 0.7 for ten predicted variables). A second model using PLS on the unfolded EEM was also conducted to improve the regression and to assess if it can handle the variability in hand. However, similar results, with a slight improvement over the MLR model, were obtained. In a non-experimental design, such variability may hinder the potentials of front-face fluorescence; however average to good MLR and PLS models were obtained, predicting the Lebanese olive oil deterioration quality parameters and fatty acid content.
