Identification of varietal and geographical origin of wines using artificial intelligence methods
Abstract The increase in wine consumption is accompanied by raising in the number of counterfeit wine products. In this regard, the task of establishing its authenticity is relevant. The purpose of this study was to build models for identifying the varietal and geographical origin of wines using mac...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-07-01
|
| Series: | Discover Food |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s44187-025-00531-2 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract The increase in wine consumption is accompanied by raising in the number of counterfeit wine products. In this regard, the task of establishing its authenticity is relevant. The purpose of this study was to build models for identifying the varietal and geographical origin of wines using machine learning methods based on the mineral composition of 426 wine samples. The elemental analysis of wines was performed using inductively-coupled plasma atomic emission and mass spectrometry methods. Discriminant Analysis showed the presence of cluster structures of wine samples relative to varietal and regional origin, which allowed the development of identification models. Classification models were built using machine learning methods: Automated neural networks, General classification and regression trees, Boosted trees classifications and regression, Random Forest, Support vector machine, k-nearest neighbor, Naïve Bayes Classifiers. At stages of solving the problem of wine identification by all methods, with the exception of Naïve Bayes Classifiers, the threshold of prediction accuracy exceeded 90%, of which the Automated neural networks method showed the best results. When determining the variety by contribution to the classification model, Rb, K, and Ca dominated, while geographical origin was dominated by Li, Mg, and Rb. It is proposed to use an ensemble approach to the classification of objects, involving the application of the voting principle, to increase the reliability of the forecast. |
|---|---|
| ISSN: | 2731-4286 |