Investigating the impact of spectral data pre-processing to assess honey botanical origin through Fourier transform infrared spectroscopy (FTIR)

Honey botanical origin is a parameter affecting its market price as certain origins are related to special organoleptic properties or potential health benefits attracting consumers’ attention. However, identifying honey botanical origin is a challenging task commonly requiring extensive high-end analysis. In this study, to address this challenge, a rapid and non-destructive attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) method was developed and special focus was paid on the spectral data pre-processing and its effect on the performance of chemometric models. Twenty-two different pre-processing methods were tested, namely, scatter correction methods, spectral derivation methods and their combinations. In each occasion, both supervised and non-supervised tools were implemented and the cross-validation parameters were used as an indicator on the efficient projection of fifty-one (n = 51) honey samples originating from 5 different botanical origins (blossom, honeydew, cotton, thyme, citrus). Importantly, combining multiplicative scatter correction followed by Savitzky-Golay first derivation is suggested as the most efficient data pre-processing method. Eventually, this data pre-processing was applied in binary models acquiring excellent recognition (87–100%) and prediction (81–100%) ability. In conclusion, the presented method set light on the undermined effect of spectral data pre-processing before the application of advanced chemometrics.