Machine Learning-based Multi-sensor Data Fusion for Quality Classification of Legen Beverage

Legen is a traditional palm sap beverage that undergoes rapid quality degradation due to spontaneous fermentation, making conventional laboratory-based assessment unsuitable for real-time monitoring. This study proposes a machine learning-based multi-sensor fusion framework for intelligent quality classification of legen beverages by integrating an electronic nose, electronic tongue, and RGB color sensor to capture aroma, taste, and visual information. The novelty of this work lies in combining heterogeneous sensor fusion with Firefly Algorithm (FA)-based sensor optimization to reduce sensor redundancy while preserving discriminative information. The results show that the optimal sensor subset, namely MQ-135, MQ-3, Copper, and RGB channels R and B, combined with the Linear Discriminant Analysis (LDA) model, achieves the best performance, with an accuracy of 91.67% and an F1-score of 91.80%. The proposed system provides a low-cost, efficient, and non-destructive solution for fermented beverage quality assessment.