<![CDATA[Portable near-infrared (NIR) spectrometers may examine samples directly on-site, speeding up data gathering. However, the NIR spectrometer has a limited wavelength, ranging from 740 to 1,070 nm, whereas previous studies used a longer wavelength. The research aims to determine the fermentation index, pH, and moisture content of fermented dry cocoa beans using a portable Near-Infrared (NIR) spectrometer. The NIR spectra were preprocessed using several methods, including the Savitzky-Golay first and second derivatives (SG1 and SG2), Linear Baseline Correction (LBC), Multiplicative Scatter Correction (MSC), comparative Partial Least Square Regression (PLSR) modeling, and Artificial Neural Network (ANN). To simulate different fermentation levels, unfermented cacao beans underwent pretreatment fermentation for durations of 0, 24, 48, and 72 hours. The prediction outcomes of ANN models, when applied to dried fermented cocoa beans with data preprocessing, offered better results in comparison to PLSR models, with strong correlation, lowest RMSEC, and highest residual predictive value. The most effective method for predicting fermentation index was ANN combined with LBC preprocessing, while optimal pH models were applied using the SG2 method. The effective moisture content models were developed using MSC preprocessing. The analytical approach of portable NIR spectroscopy produced rapid and accurate results to determine the quality of ground-dried cacao bean fermentation.]]>
Copyrights © 2025
