Development of a Decision Support System for Wheat Quality Prediction based on Expert Knowledge and Supervised Machine Learning
2023
Munch, Mélanie | Baudrit, C | Saulnier, Luc | Kansou, Kamal | Institut de Mécanique et d'Ingénierie de Bordeaux (I2M) ; Université de Bordeaux (UB)-Institut Polytechnique de Bordeaux-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement (INRAE)-Arts et Métiers Sciences et Technologies | Unité de recherche sur les Biopolymères, Interactions Assemblages (BIA) ; Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement (INRAE) | ANR-20-CE21-0008,EVAGRAIN,Des outils intelligents pour une utilsation agile du blé(2020)
International audience
Mostrar más [+] Menos [-]Inglés. Due to growing constraints (global warming, customers’ expectations, new agricultural practices), wheat harvest today is characterized by its wide diversity. As a result, tools and/or methods are no longer suitable and not selective enough for assessing relevant wheat quality. The development of new quality indicators able to predict the many possible uses of wheat is becoming capital. This work is part of the ANR Evagrain project, whose ambition is to design a decision support system able to predict a wheat’s quality . In this context, quality focuses on the use value, i.e. the extent to which a given wheat will produce satisfactory results for its assigned purpose. With this in mind, machine learning algorithms have been studied due to their flexibility useful to face the diversity of wheats and uses. From a database of analytical tests, representing the potential predicting criteria, models have been trained to predict given quality profiles validated by experts . The idea is to provide experts a tool able to predict a wheat’s given usage from measurable characteristics.Support Vector Machine (SVM) algorithm presents the best prediction’s accuracy. By analyzing the impact of each feature of the model in order to give interpretability and explainability about its predictions, the alveogram parameters (P and Ie), the gluten Index and the protein content play major roles. This highlights potential parameters that can inform the most an expert about a wheat’s best use. This can also be extended to test new potential predicting parameters.
Mostrar más [+] Menos [-]Palabras clave de AGROVOC
Información bibliográfica
Este registro bibliográfico ha sido proporcionado por Institut national de la recherche agronomique