Near-Infrared Spectroscopy-Based Discriminant Analysis for the Classification of Coffee Quality in Dry Parchment and Green Coffee.

This study evaluates the potential of near-infrared spectroscopy (NIRS) combined with discriminant analysis to classify coffee quality based on sensory defects in dry parchment coffee (DPC) and green coffee. Spectral data were used to develop classification models, which were validated using both cross-validation and independent external datasets. Model performance was assessed using classification accuracy and Cohen's kappa coefficient. The results demonstrate high classification accuracy for DPC (93.5%), with a Kappa coefficient indicating almost perfect agreement (κ = 0.90). In contrast, green coffee showed lower predictive performance (82.4%) and moderate agreement (κ = 0.55), reflecting the greater physicochemical complexity of this matrix. Importantly, the findings demonstrate that coffee quality can be reliably classified at the dry parchment stage, enabling early quality assessment without additional processing steps. This represents a significant advancement compared to previous studies, which have mainly focused on green or roasted coffee. Overall, these results highlight the potential of NIRS as a rapid, non-destructive, and objective tool for coffee quality assessment, with strong applicability in quality control and decision-making processes along the coffee production chain.