Detection of Jaggery Syrup Adulteration in Apis mellifera Honey Using a Multispectral Sensor Coupled with Machine Learning

Abstract

Honey adulteration with low-cost sugar syrups, such as jaggery syrup, invert sugar syrup, and corn syrup, has become a major concern affecting honey quality and purity. The present investigation aimed to develop a rapid, non-destructive method for detecting jaggery syrup (JS) adulteration in Apis mellifera honey using an AS7341 multispectral sensor coupled with machine-learning techniques. Pure honey was adulterated with jaggery syrup at five concentrations (10, 20, 30, 40, and 50%, w/w), and its physico-chemical properties were determined. Significant differences (P < 0.01) were observed in moisture content, electrical conductivity, colour, and pH between pure honey and jaggery syrup, whereas specific gravity, total ash, and acidity expressed as formic acid showed no significant variation. The AS7341 multispectral sensor recorded variations in spectral intensity with increasing adulteration levels, particularly in the F1, F2, F3, and near-infrared (NIR) channels, in which spectral intensity progressively declined as the level of jaggery syrup increased. These spectral intensity data were used to develop Random Forest (RF) and Artificial Neural Network (ANN) classification models. The RF model achieved 100% accuracy, precision, recall, and F1-score for both the training and testing datasets, correctly classifying all adulteration levels. In contrast, the ANN model achieved 33.33% accuracy, 20.00% precision, 33.33% recall, and 22.22% F1-score, indicating limited classification capability under the present experimental conditions. The developed methodology may have practical applications in honey quality assurance and purity assessment.