| dc.description.abstract | This work presents the use of near-infrared (NIR) spectroscopy as a fast and eco-friendly alternative to the conventional methods currently used for the quality control of cocoa beans. A key feature of this study was that the multivariate calibration models were built from the spectra of 81 samples of commercial cocoa beans from different producing regions and sampled over the period of 1 year. This aspect is crucial to demonstrate the feasibility of NIR for predicting chemical parameters of cocoa beans, since it provides a realistic variability in the calibration models. PLS regression models were constructed from the near-infrared diffuse reflectance spectra, allowing the prediction of moisture, pH, acidity, fat, shell content, protein, total phenolic compounds, caffeine, and theobromine through direct analysis of samples milled and sieved without any additional preparation. All of these parameters were predicted with adequate coefficient of determination values (R2), which ranged from 0.67 to 0.89 and relative errors smaller than 10.2%, which is quite suitable when compared to the coefficients of variation of the reference methods. Thus, the results demonstrated that the use of NIR combined with chemometrics is effective and recommended for the quality control of commercial cocoa beans. | |
