Intelligent Quality Assessment of Fruit-Vegetable Enzyme Fermentation via Multisensor Fusion and Machine Learning

Consistent quality assurance and shelf-life prediction of enzyme beverages face challenges from rapidly changing physicochemical conditions during fermentation. This study proposes an end-to-end framework integrating data from 11 common sensors (pH, alcohol, SO₂, sugars) for dual tasks: pass-fail grading and shelf-life estimation. Five machine learning algorithms were tested on 6497 Wine-Quality records split into training, validation, and test sets. Random Forest outperformed others, achieving 0.919 area under the receiver operating characteristic curve (AUROC) and 0.932 recall for classification, the 0.394 mean absolute error (MAE), 0.301 root mean squared error (RMSE), and highest R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> (0.594) for regression. Analysis showed minimal bias and tightest error distribution. Shapley (SHAP) values identified alcohol, volatile acidity, and free SO₂ as key factors aligning with fermentation chemistry.