Machine Learning and SHAP Feature Analysis: Classification Model for Aroma Components in Green Plum Wine

This study employed GC–MS, sensory evaluation, and odor activity value (OAV) analysis combined with machine learning to differentiate the volatile flavor profiles of fermented green plum wines. Most green plum wines contained high levels of esters, with ethyl benzoate (up to 4820.53 μg/L), ethyl caprylate (up to 2640.83 μg/L), and benzaldehyde (up to 3432.96 μg/L) being the major contributors. Among six compared algorithms, fuzzy c-means clustering performed best, distinguishing three distinct flavor categories, while a decision tree model achieved 95.13% classification accuracy. SHAP analysis identified ethyl caprylate, benzyl acetate, and phenethyl acetate as the most influential volatile compounds. These results demonstrate that machine learning provides an effective approach for flavor classification and interpretation in fruit wines.