Electronic Nose and Optimized Machine Learning Algorithms for Noninfused Aroma-Based Quality Identification of Gambung Green Tea

Tea is a widely consumed beverage globally, significantly impacting taste, aroma, and consumer satisfaction. Traditional methods of assessing tea quality rely on subjective sensory evaluation by expert panels, which can be time-consuming, expensive, and prone to biases. In recent years, electronic nose (e-nose) technology has emerged as a promising and objective approach to tea quality assessment. This study investigates the application of e-nose technology for the automated and objective assessment of Gambung green tea quality, especially for noninfused (dry) tea aroma. Using noninfused tea has advantages because it does not require a long preparation procedure such as brewing and does not damage the tea sample, so it can be used repeatedly. However, this also has its challenges due to the less strong tea aroma making it more difficult to identify. Machine learning algorithms and hyperparameter optimization (HPO) were utilized to evaluate their performance in classification and regression tasks. The k-nearest neighbor (k-NN) algorithm achieved the highest accuracy and provided reliable confusion matrix results. Moreover, the k-NN algorithm outperformed other approaches in predicting organoleptic scores, with the highest
${R}$
-squared score and the lowest mean squared error values. In summary, this study demonstrates the potential of e-nose technology coupled with optimized machine learning algorithms for the noninfused aroma-based identification of Gambung green tea. However, improvements are required to enhance the performance of organoleptic score predictions. This study contributes to the understanding and application of e-nose technology in tea quality assessment, opening ways for further progress in this area.