Ultraviolet–visible spectroscopy combined with machine learning as a rapid detection method to the predict adulteration of honey

Abstract

Honey is often adulterated with inexpensive and artificial sweeteners. To overcome the time-consuming honey adulteration tests, which require precision, chemicals, and sample preparation, it is needful to develop trustworthy analytical methods to assure its authenticity. In the present study, the potential of ultraviolet–visible spectroscopy (UV–Vis) in predicting the sucrose content was evaluated by using Support Vector Regression (SVR) and Partial Least Square Regression (PLSR). To predict the sucrose content based on diagnostic wavelengths, a Point Spectro Transfer Function (PSTF) was evaluated using Multiple Linear Regression (MLR). For this purpose, the spectra of authentic (n = 12), commercial (n = 12), and adulterated (n = 16) honey samples were recorded. Four distinguished wavelengths from correlation analysis between sucrose content and spectra absorption were 216, 280, 316, and 603 nm. The SVR performed better calibration model than the PLSR estimations (RMSE = 0.97, and R2 = 0.98). The predictive models result revealed that both models had high accuracy for the sucrose content estimation. This study proved that UV–Vis spectroscopy provides an economical alternative for the rapid quantification of adulterated honey samples with sucrose.

Keywords

• Adulteration
• Authentication
• Honey fraud
• Machine learning
• Spectral pre-processing