WineRec – AI: Software solution for wine authenticity control

Authors: Ariana Raluca Hațegan, Dana Alina Măgdaș, Ana Camelia Groșan

Keywords: wines, authenticity, machine learning, spectroscopic methods

Applications

Assessing the authenticity of beverages in an efficient and rapid manner is an international concern. In the case of wines, due to the many factors influencing its composition, evaluation includes, among others, the analysis of grape variety, geographical origin or year of production.

While most of the widely recognized techniques for wine authenticity control are costly and require specialized knowledge, the developed web application is based experimental data sets acquired through the application of different spectroscopic techniques (1H-NMR and Raman), allowing the implementation of rapid methods for wine fingerprinting.

The software solution is therefore an easy tool in the field of food and beverage authentication and can be easily used by people who do not have advanced knowledge in experimental data processing.

Innovative aspects

The web application has a high degree of novelty at a national and international level, as there is currently no similar tool known to exist that allows the development of new models for wine origin recognition based on the use of rapid analytical techniques, such as 1H-NMR or Raman, in association with machine learning algorithms.

The main innovative aspect of the software solution corresponds to its intrinsic experimental data processing procedure, which allows the development of efficient wine differentiation models, depending on the user’s options for input data type, spectral domain of interest, classification criteria (geographical origin, variety, year of production) or machine learning algorithm, among others.

The application has a database containing the 1H-NMR and Raman spectra corresponding to a set of more than 100 authentic wine samples, as well as various optimized recognition models based on machine learning algorithms to identify their variety, geographical origin and year of production.

Technology

The proposed web application for wine authenticity control was developed using Django, an open-source web development framework based on the Python programming language.

In order to meet the expectations of potential users, the WineRec – AI information system includes the following key functionalities:

(i) Authentication mechanisms to ensure secure access and protection of experimental data,
(ii) tools for efficient database management and
(iii) tools for developing and applying predictive models.

Advantages

  • possibility to use several types of experimental data (1H-NMR or Raman) to identify the origin of wines;
  • application and development of wine origin prediction models based on advanced data processing methods (machine learning algorithms) under an intuitive and user-friendly interface that does not require computer expertise;
  • automatic optimization of recognition models by determining the optimal values of the hyper-parameters specific to the selected machine learning algorithm;
  • efficient management of experimental data and the possibility to use them for the development of wine origin prediction models in a centralized computer system, allowing remote access and sharing of information among multiple users.