5-Hydroxymethyl furfural determination in Italian honeys by a fast near infrared spectroscopy

The determination of 5‑hydroxymethylfurfural (HMF) content in honey samples of different botanical origin by a NIR-chemometrics-based approach is reported. Spectral regions, statistical models, scatter and derivative correction together with smoothing pre-treatment were examined and selected on the basis of correlation coefficient and root mean square errors for both the calibration (R2c and RMSEC) and cross-validation (R2v and RMSECV). Prediction ability was tested on an external set of honey samples and the best results were achieved with a Partial Least Square regression approach coupled with a multiplicative signal correction of the scattering in the spectral region 4252-4848 cm1. The correlation coefficient in prediction (R2 = 0.98) as the residual predictive deviation (RPD = 3.3) suggest that the proposed model proves to be a powerful tool to fulfil quality goals in such a complex matrix as honey.