Botanical Origin Assessment of Honey Based on ATR-IR Spectroscopy: A Comparison between the Efficiency of Supervised Statistical Methods and Artificial Intelligence

Abstract

Food authenticity control represents a constant concern nowadays, and against this background, new means of food fraud detection are developed by research and control laboratories. Among the most accessible analytical methods in this regard, attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy proved to be an effective tool, being rapid, cost-effective, and not requiring solvent use. However, the generated experimental data need to be further processed in an efficient manner in order to be able to accurately assess the authenticity of a certain product. The temptation to pass some more available honey varieties as rarer ones might exist and in order to detect these types of miss labeling, we proposed in this study the development of new recognition models based on supervised chemometric models and artificial intelligence. In this way a comparison between the models’ capabilities constructed based on the association between ATR-IR spectroscopy with partial least squares discriminant analysis (PLS-DA) and support vector machines (SVM), respectively, was performed. The most efficient models for the individual botanical differentiation were developed by applying SVM on the significant spectral markers, determined through a supervised method.