Scientists at the University of Geneva have developed an artificial intelligence algorithm that can track the origin of wine based on its chemical composition. This breakthrough method could be a valuable tool in the fight against wine fraud. Led by Professor Alexander Pozh, the researchers used machine learning to analyze the concentrations of various compounds in wine, allowing them to not only determine the region where the grapes were grown but also the specific winery responsible for its production.
The study focused on 80 wine samples collected over a 12-year period from seven different sections in the Bordeaux region. By employing gas chromatography, the team was able to analyze the overall chemical composition of the wine rather than examining individual compounds, thus creating a unique “signature” for each wine.
What sets this algorithm apart is its ability to present results on a two-dimensional grid, grouping wines with similar “signatures” together. This discovery has immense potential in combating fraudulent wine production, which causes billions of euros in damage each year.
However, while the algorithm successfully distinguishes between different wineries with 99% accuracy, it struggles to determine the year of production, achieving only a 50% accuracy rate. The findings of this study, set to be published in the journal comunications chemistry, underscore the significant role that machine learning can play in the food and agriculture industry.
Moreover, in addition to detecting counterfeit wines, this method can also be utilized to monitor the quality of wine throughout the production process and optimize blending techniques. This breakthrough has the potential to make the production of high-quality wines more cost-effective and efficient.