PPCCA/Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627, 31270-901 Belo Horizonte, MG, Brazil.
DQ/Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627, 31270-901 Belo Horizonte, MG, Brazil.
Food Chem. 2018 Apr 15;245:1052-1061. doi: 10.1016/j.foodchem.2017.11.066. Epub 2017 Nov 20.
Sensory (cup) analysis is a reliable methodology for green coffee quality evaluation, but faces barriers when applied to commercial roasted coffees due to lack of information on roasting conditions. The aim of this study was to examine the potential of mid-infrared spectroscopy for predicting cup quality of arabica coffees of different roasting degrees. PCA analysis showed separation of arabica and robusta. A two-level PLS-DA Hierarchical strategy was employed, with coffee being classified as high or low quality in the first level and then separated according to cup quality in the second level. Validation results showed that the second level models exhibited 100% sensitivity and specificity in the training sets. For the test set, sensitivity ranged from 67% (rio zona) to 100% (soft) while specificity ranged from 71% (rio) to 100% (rioysh, hard). Thus, the proposed method can be used for the quality evaluation of arabica coffees regardless of roasting conditions.
感官（杯）分析是评估绿色咖啡质量的可靠方法，但由于缺乏烘焙条件信息，应用于商业烘焙咖啡时存在障碍。本研究旨在探讨中红外光谱法在预测不同烘焙程度阿拉比卡咖啡杯测质量方面的潜力。PCA 分析显示阿拉比卡和罗布斯塔的分离。采用两级 PLS-DA 分层策略，在第一级将咖啡分为高质量和低质量，然后在第二级根据杯测质量进行分离。验证结果表明，第二级模型在训练集中的灵敏度和特异性均为 100%。对于测试集，灵敏度范围为 67%（里约扎纳）至 100%（柔和），特异性范围为 71%（里约）至 100%（里约伊什，硬）。因此，该方法可用于评估阿拉比卡咖啡的质量，而不受烘焙条件的影响。
