Colorimetric sensor array combined with machine learning for discriminating the geographical origins and grape varieties of Chinese red wines

The volatile compounds present in wines are the main determinants of aroma performance, which can be used as chemical markers to differentiate between wine samples. In this work, a colorimetric sensor array (CSA) combined with machine learning was developed for rapid and non-destructive discrimination of the geographical origins and grape varieties of Chinese red wines. Twelve chemically responsive dyes were selected to capture the volatile profiles of wine samples, and the resulting colorimetric responses were converted into multidimensional features using image analysis. Principal component analysis (PCA) was applied for dimensionality reduction, and three machine learning models including random forest (RF), K-nearest neighbor (KNN), and support vector machine (SVM) were constructed for classification. It was demonstrated that the SVM model exhibited superior classification performance compared with the other models, achieving recognition rates of 95.0% for origins and 95.5% for varieties in the prediction set. This study showed that the combination of CSA and machine learning is an effective, rapid, and accurate method for wine authenticity verification, offering a promising approach for quality control in the wine industry. • A colorimetric sensor array (CSA) was constructed to discriminate Chinese red wines. • PCA was applied to reduce data dimensionality and visualize of the digital fingerprints of CSA. • Machine learning algorithms were used to classify the origins and varieties of wines based on CSA. • SVM had the best performance for discriminating both origins and varieties of Chinese red wines. • CSA integrated with machine learning is a promising method for wine authenticity verification