Honey Differentiation Using Infrared and Raman Spectroscopy Analysis and the Employment of Machine-Learning-Based...

Area of Science:

• Analytical Chemistry
• Food Science
• Spectroscopy

Background:

• Rising concerns about honey adulteration and mislabeling necessitate reliable authentication methods.
• European directives encourage rapid, cost-effective analytical techniques like vibrational spectroscopies for honey analysis.
• Varietal-dependent composition influences the spectral fingerprint of different honey types.

Purpose of the Study:

• To identify key vibrational bands in FT-Raman and ATR-IR spectra for major Transylvanian honey varieties.
• To develop and evaluate a novel methodology for honey authentication using combined vibrational data and machine learning.
• To assess the model's accuracy in distinguishing honey from a narrow geographical region and specific floral sources.

Main Methods:

• Utilized Fourier Transform Infrared (FT-IR) and FT-Raman spectroscopy to analyze acacia, honeydew, and rapeseed honey.
• Developed a new data processing methodology combining IR and Raman spectral information.
• Applied machine learning algorithms, specifically the Trilayered Neural Network, for data analysis and model development.

Main Results:

• The developed model accurately distinguished honey varieties based on their unique spectral fingerprints.
• The Trilayered Neural Network achieved 85.2% accuracy on training and 93.8% on testing datasets for regional differentiation.
• Acacia honey was differentiated from fifteen other sources with 87% accuracy.

Conclusions:

• The proposed methodology effectively authenticates honey based on vibrational spectroscopy and machine learning.
• This approach offers a reliable tool for honey label control and enhancing food safety.
• The study demonstrates the potential for precise geographical and floral origin determination of honey.