Robust Honey Adulteration Detection Through Ensemble Techniques with Hyperspectral Analysis

Honey adulteration detection involves the use of Ensemble Machine Learning (EML) techniques to classify raw hyperspectral honey data. Adulterants are substances added to honey to increase the volume and utility of honey. Sugar syrup, corn syrup, and glucose are typical adulterants that hinder the purity and nutritional value of honey. Hyperspectral imaging (HSI) provides detailed spectral data with high spatial resolution. A standard hyperspectral Honey dataset with 128 features is considered for the study. The primary phase consists of the pre-processing of the dataset. The EML techniques integrate individual models to obtain a more precise and generalized predictive model. The pre-processed data was subjected to classification by eight different EML models (four Bagging (Bagging Decision Tree (BDT), Random Forest (RF), Random Subspace (RS), and Random Patches (RP)) and four Boosting (Gradient Boosting (GB), Light Gradient Boosting (LGB), Categorical Boosting (CB), and eXtreme Gradient Boosting (XGB)). EML model outcomes were evaluated using various performance measures such as Accuracy, Precision, Recall, F1-score, Cohen's Kappa, Matthew's Correlation Coefficient (MCC), and Area Under the Receiver Operating Characteristic Curve (AUC-ROC). The XGB Boosting EML technique gives an outstanding performance with an accuracy value of 91.97%, precision value of 91.89%, recall value of 91.92%, f1-score value of 91.98%, kappa value of 91.71%, MCC value of 91.87%, and AUCROC value of 92.99%. The findings indicate that ensemble methods significantly improve the identification of honey adulteration when compared to individual models. Boosting techniques, specifically, display exceptional accuracy and resilience in classification assignments.