Infrared spectroscopy and genetic algorithm based spectral feature selection for rapid honey adulteration detection

This study involves the authentication of pure honey samples against various adulterants, including those derived from C3 plants (rice syrup) and C4 plants (sugar syrup, corn syrup, and jaggery syrup). A total of thirty varieties of honey samples, along with four adulterants, were analyzed using both NIR and FTIR spectroscopy to generate a comprehensive spectral dataset for chemometric-based authentication. The comparative application of NIR and FTIR aimed to evaluate their relative performance in terms of spectral discrimination, classification accuracy, and practical suitability for adulteration detection. To validate the authenticity of the honey samples, stable carbon isotope ratio analysis (δ¹³C) was performed using Elemental Analyzer and Liquid Chromatography-Isotope Ratio Mass Spectrometry (EA/LC-IRMS), which also served as a reference standard for calibration models. The spectral dataset was pre-processed followed by the relevant spectral features selection using a combination of PCA loadings and Genetic Algorithm (GA), enhancing model interpretability and classification performance. For qualitative prediction, classification models (PCA + K-medians, SVM, and LDA) achieved up to 100% accuracy across all adulterants, with NIR yielding 100% accuracy for SVM and LDA, and FTIR showing 100% accuracy for LDA and K-medians. For quantitative prediction, the best regression results were obtained using SVR for NIR (R² = 0.9987, RMSE = 1.265% for jaggery syrup) and PCR for FTIR (R² = 0.9982, RMSE = 1.475%). The study demonstrates that infrared spectroscopy, combined with GA-based feature selection and chemometric modeling, provides a robust and scalable predictive framework for honey adulteration detection.