Aflatoxins contamination in Pistachio has been known as a serious problem for export market. Considering the difficulties of current experimental techniques for aflatoxins detection in pistachio (complication, time consuming and experience required), Raman spectroscopy combined with linear discriminant analysis technique was used to classify intact and aflatoxins contaminated pistachio samples (in concentrations of 100 and 20 ng/g). Raman spectra were analyzed in the spectral region of 848-2000 cm-1. The effect of the preprocessing methods such as smoothing, normalization, the first and second derivatives, detrending, standard normal variate and multiplicative scatter correction and their combination on classification results was also studied. To evaluate the classification model, Sensitivity, harmonic mean of sensitivity and precision (Fβ) of the classification results (confusion matrix) were calculated. Results indicated that Raman spectroscopy combined with linear discriminant analysis technique have potential to classify pistachio samples with and without aflatoxins. Moreover, all preprocessing methods reduced the average sensitivity and the average of Fβ, except Savitzky-Golay (SG) smoothing method (Fβ= 0.73, sensitivity= 72.22). The maximum value of Fβ= 0.86 and sensitivity= 83.33 for intact samples classification using normalization or SG + normalization preprocessing methods were achieved.
