Predicting Green Tea Organoleptic Scores Based on Electronic Nose Dataset using Boosting Algorithms

Traditional tea quality assessment relies on subjective evaluations from human tasters, leading to inconsistencies and potential biases. This study explores the potential of Electronic Nose (E-Nose) technology combined with machine learning to overcome these limitations. We leverage data from the Gambung green tea and quinine research center, including organoleptic scores. Hyperparameter optimization (HPO) is employed on various algorithms, including XGBoost, K-Nearest Neighbors, Decision Tree, and AdaBoost. The results demonstrate XGBoost's superiority, achieving excellent R-squared (0.97) and Mean Squared Error (MSE) (0.03) values, indicating its effectiveness in modeling tea quality from aroma data. Scatter plot analysis reveals consistent patterns in XGBoost predictions, while K-Nearest Neighbors shows promising accuracy (R-squared of 0.89). Decision Tree results exhibit significant variation, and AdaBoost predictions show greater dispersion. These findings establish XGBoost as the most effective algorithm for predicting tea quality based on aroma, including organoleptic scores. In conclusion, this study underscores the value of employing E-Nose and machine learning for objective tea quality assessment through aroma analysis, paving the way for a more consistent and reliable approach.