Artificial Intelligence Applied to Honey ¹³C NMR Data: A New Path for Honey Recognition

Besides food authentication control based on acknowledged analytical methods, nowadays there is a continuous attempt to develop new approaches for unambiguously differentiating distinct food commodities. In this regard, current tendencies involve Artificial Intelligence (AI) to develop robust recognition models, especially when subtle differences in the experimental data exceeding human capabilities need to be learned for a final verdict. This study proposes the pairing of AI and ¹³C NMR spectra to develop honey geographical and botanical discrimination models. To test the suitability of this approach, a set comprising more than 100 honey samples from Romania and France was employed. Because ¹³C NMR generates a large variable set, the development of reliable classification tools was achieved when the input data space was limited to the most relevant features. Through this approach, highly effective models for the geographical and botanical differentiation of honey, having accuracy scores greater than 97% in cross-validation, have been constructed.