Abstract
Accurate identification of liquor vintage is crucial for ensuring product authenticity and optimizing market value, as the price and sensory quality of liquor increase with age. Traditional sensory evaluation by sommeliers is inherently limited by subjectivity, physiological fatigue, and inconsistency, posing challenges for reliable large-scale quality assessment. To address these limitations, this study introduces an innovative homemade electronic tongue (ET) system integrated with machine learning and deep learning algorithms for rapid and precise vintage identification. The ET system, consisting of six metallic electrodes and a MEMS-based temperature sensor, successfully discriminated five consecutive liquor vintages produced at one-year intervals. Using Support Vector Machine (SVM) and Random Forest (RF) algorithms, classification accuracies of 91.0% and 78.0% were achieved, respectively. Remarkably, the proposed one-dimensional convolutional neural network (1D-CNN) model further improved the recognition accuracy to 94.0%, representing the highest reported performance for ET-based vintage prediction to date. The findings demonstrate that the integration of multi-electrode electrochemical sensing with artificial intelligence enables objective, reproducible, and high-throughput evaluation of liquor aging characteristics. This approach provides a scientifically robust alternative to human sensory analysis, offering significant potential for counterfeit detection, liquor authentication, and the broader assessment of food and beverage quality within molecular sensing frameworks.