Comparison of different classification methods for analyzing fluorescence spectra to characterize type and freshness of olive oils

The potential of synchronous fluorescence spectroscopic measurements for the classification of olive oil samples with respect to type (extra virgin/refined) and shelf-life condition (expired/non-expired) was examined. Data multidimensionality was reduced by application of the PCA method. Six classification methods, namely, LDA, QDA, RDA, KNN, SVM, and RF, were used to analyze fluorescence data. Classification accuracy was employed to evaluate the performance of these methods for the classification of olive oil samples. The results indicate that fresh and expired olive oils produced with different methods, extra virgin (cold pressed) and refined, are characterized by different fluorescence spectra. KNN and SVM outperform the other classification methods in distinguishing olive oils. The lowest classification error rates were obtained with these two classification methods for measurements at the wavelength interval of 30nm, and equaled 5.4, 5.7 and 2.9, 3.5% for the four and three classes' models, respectively.