Characterization and classification of odorous raw milk: Volatile profiles and algorithm model perspectives

Odorous raw milk poses a growing threat to dairy product quality, negatively impacting both producers and consumers. However, methods for effectively identifying odorous raw milk have not been systematically established. In this study, 30 raw milk samples (RMSs) collected from different pastures were classified into 18 fresh RMSs and 12 odorous RMSs through sensory evaluation. Quantitative datasets of volatile compounds in RMSs were obtained using HS-SPME–GC–MS. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) models identified nine key differential volatile compounds distinguishing fresh from odorous RMSs. Among these, hexanal and octanal were identified as potent odorants contributing to the off-odor in raw milk. Based on nine key difference volatile compounds, support vector machine (SVM), multilayer perceptron (MLP), and random forest (RF) algorithmic models were successfully constructed to classify RMSs to identify odorous RMSs. All models achieved classification accuracy exceeding 0.9, with the RF model performing the best, achieving an accuracy of 1.0. This work provides a reference and available workflow for identifying and labeling odorous RMSs.