Alternating conditional expectation (ACE) algorithm and supervised classification models for rapid determination and classification of the adulterated cinnamon samples using diffuse reflectance FT-IR spectroscopy

• Published in: Journal of food composition and analysis Vol. 131; p. 106226
• Main Authors: Hajiseyedrazi, Zahra S., Khorrami, Mohammadreza Khanmohammadi, Mohammadi, Mahsa
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.07.2024
• Subjects: adulterated products; algorithms; black pepper; calibration; Chemometric data analysis; chemometrics; Cinnamomum aromaticum; Cinnamomum verum; cinnamon; Cinnamon adulterations; food composition; Fourier transform infrared spectroscopy; MIR spectroscopy; PLS-DA; reflectance; regression analysis; risk; rPCA-ACE; SVM-DA; MIR spectroscopy; SVM-DA; PLS-DA; rPCA-ACE; Cinnamon adulterations; Chemometric data analysis
• ISSN: 0889-1575; 1096-0481
• DOI: 10.1016/j.jfca.2024.106226

Cinnamon authentication is an important subject in the field of spice adulteration studies due to its widespread applications in the food and pharmacy industries. Rapid, non-destructive, and smart approaches lead us to the best results. Cinnamomum verum (C. verum), as the high-priced species of cinnamon, is at risk of mixing with Cinnamomum cassia (C. cassia), black pepper, and clove. Thus, C. verum was selected as our analyte, and the rest constructed the adulterants group. In this work, diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) associated with chemometric procedures was utilized for the determination and classification of adulterated cinnamon samples. Two problems were defined: a) classification of multi-adulterated samples into three levels of adulteration (i.e. low, medium, and high) and b) detection of adulterated samples by predicting concentration values. Partial least squares-discrimination analysis (PLS-DA) and Support vector machine-discrimination analysis (SVM-DA) were used for classification problems to compare the performances of linear and non-linear models. Additionally, the robust principal component analysis-alternating conditional expectation (rPCA-ACE) model, as the representative of robust models, was utilized for regression to predict the concentration value of the analyte (C. verum) in samples. In fact, in the regression model, the adulteration of samples was determined using the concentration values of the analyte in FTIR spectra. In the classification models, statistical parameters such as accuracy, precision and error rate were calculated. The accuracy of PLS-DA and SVM-DA for the calibration set is 0.960 and 0.972, respectively. These results show better performance of SVM in the classification section. The rPCA-ACE model in the regression showed good efficiency performance with an R2 of calibration of 0.838. Therefore, the results demonstrate that DRIFTS coupled with chemometric data analysis would be a capable strategy for the determination and classification of adulterated cinnamon samples. [Display omitted] •DRIFTS with chemometrics were used for determination and classification of adulterated cinnamon samples.•PLS-DA and SVM-DA were utilized as the classification models.•The results showed better performance of SVM-DA in the classification section.•rPCA-ACE as a multivariate calibration model was used.