Raman spectroscopy combined with machine learning algorithms to detect adulterated Suichang native honey

Area of Science:

• Analytical Chemistry
• Food Science
• Spectroscopy

Background:

• Zhejiang Suichang native honey is a popular product protected under China's National Geographical Indication Agricultural Products.
• Adulteration of honey, especially at low concentrations, poses a significant challenge to quality control and consumer trust.

Purpose of the Study:

• To develop and validate a method for accurately detecting low-concentration adulteration in Suichang native honey.
• To evaluate the effectiveness of Raman spectroscopy combined with machine learning algorithms for honey authentication.

Main Methods:

• Collected Suichang native honey samples for adulteration analysis.
• Applied Savitzky-Golay smoothing and Partial Least Squares (PLS) for spectral data compression.
• Utilized Support Vector Machine (SVM), Probabilistic Neural Network (PNN), and Convolutional Neural Network (CNN) for classification modeling.

Main Results:

• Achieved classification accuracies of 100% (SVM), 100% (PNN), and 99.75% (CNN) for pure and adulterated honey samples.
• Demonstrated successful feature selection based on PLS contribution rate for effective analysis.
• Confirmed the capability of the combined method in identifying even low levels of honey adulteration.

Conclusions:

• Raman spectroscopy coupled with machine learning algorithms offers a powerful and accurate approach for detecting honey adulteration.
• The proposed method holds significant potential for ensuring the authenticity and quality of geographical indication honey products.
• This technique provides a reliable tool for food safety and quality assurance in the honey industry.