Probabilistic modeling of the quality of a vineyard and deterministic modeling of the quality of a wine: first approaches
Abstract
Objectively evaluating the quality of a vineyard is not always simple. We designed a probabilistic model that takes into account the parameters defining the status of the vineyard with their associated interactions. It is based on a Bayesian network, including an inference engine and software. Data were collected from vine-growing experts. For this purpose, 661 grape varieties were evaluated by experts and data for these varieties were included in the system. No such model has been developed before. This system should ease assessments of the likely impact of the choices and decisions of vine growers on the quality of new vineyards to be planted in any part of the world.
Predicting the quality of a wine is a complex task given the large number of factors to be taken into account simultaneously and their interactions. This work is a first deterministic approach to assess the quality of a wine. It uses a very important dataset collected for decades in a vineyard associated to the expertise of the wine maker. In this approach, we used the methods of linear algebra to represent the quality of the wine through several sub-models. Created for the situations studied (Châteauneuf du Pape, Côtes du Rhône, Rasteau), this expert model is not directly applicable to other situations but could be used as a reference for further studies.
Issue: GiESCO 2015
Type: Extended abstract
Authors
1 INRA, UMR 1083, Science for Enology, bat. 28, 2 Place Viala, 34060 Montpellier, France
2 Université de Rennes II, CS24, 307, rue Henri Le Moal, 35043, Rennes, France
3 INRA-SupAgro UMR 1034, AGAP,Bt 21, 2 Place Viala, F34060 Montpellier, France
4SupAgro, 2, Place Viala, 34060 Montpellier, France
Contact the author*
Keywords
vineyard quality, bayesian network, expert data, linear model, wine quality