Unbiased Characterization and Classification of Coffee

Coffee quality assessment traditionally relies on certified sensory specialists who are not easily accessible. The reliance on subjective evaluations can lead to inconsistencies and inaccuracies in coffee assessment, making it difficult to achieve reliable and standardized evaluations. This study leverages machine learning to objectively classify brewed coffee by brand, class, and roast level using physicochemical characterization. 112 brewed coffee samples from six different coffee brands ranging across commodity, specialty, and premium classifications, at varying roast levels, were subject to pH, color, turbidity, potassium ion, and total dissolved solid measurements. We utilized Support Vector Machine (SVM), Decision Tree, and Logistic Regression models to predict coffee brand, class, and roast level. SVM, enhanced by hyper-parameter tuning, achieved 100\% accuracy using 10-fold cross-validation, outperforming Decision Tree and Logistic Regression models. Key features essential to high-accuracy classification were pH, turbidity, color, and potassium ion content. Our findings demonstrate that machine learning models can accurately classify quality, brand, and roast level, offering a reliable and cost-effective alternative to traditional subjective methods. This approach may benefit manufacturers in maintaining consistent quality and aid consumers in identifying their preferred coffee profiles.