In this paper, we describe a new computational methodology to select the best regression model to predict a numerical variable of interest Y and to select simultaneously the most interesting numerical explanatory variables strongly linked to Y. Three regression models (parametric, semi-parametric and non-parametric) are considered and estimated by multiple linear regression, sliced inverse regression and random forests. Both the variables selection and the model choice are computational. A measure of importance based on random perturbations is calculated for each covariate. The variables above a threshold are selected. Then a learning/test samples approach is used to estimate the Mean Square Error and to determine which model (including variable selection) is the most accurate. The R package modvarsel (MODel and VARiable SELection) implements this computational approach and applies to any regression datasets. After checking the good behavior of the methodology on simulated data, the R package is used to select the proteins predictive of meat tenderness among a pool of 21 candidate proteins assayed in semitendinosus muscle from 71 young bulls. The biomarkers were selected by linear regression (the best regression model) to predict meat tenderness. These biomarkers, we confirm the predominant role of heat shock proteins and metabolic ones.

The beef industry is organized around different stakeholders, each with their own expectations, sometimes antagonistic. This article first outlines these differing perspectives. Then, various optimization models that might integrate all these expectations are described. The final goal is to define practices that could increase value for animal production, carcasses and meat whilst simultaneously meeting the main expectations of the beef industry. Different models previously developed worldwide are proposed here. Two new computational methodologies that allow the simultaneous selection of the best regression models and the most interesting covariates to predict carcass and/or meat quality are developed. Then, a method of variable clustering is explained that is accurate in evaluating the interrelationships between different parameters of interest. Finally, some principles for the management of quality trade-offs are presented and the Meat Standards Australia model is discussed. The “Pareto front” is an interesting approach to deal jointly with the different sets of expectations and to propose a method that could optimize all expectations together.

The beef cattle industry is facing multiple problems, from the unequal distribution of added value to the poor matching of its product with fast-changing demand. Therefore, the aim of this study was to examine the interactions between the main variables, evaluating the nutritional and organoleptic properties of meat and cattle performances, including carcass properties, to assess a new method of managing the trade-off between these four performance goals. For this purpose, each variable evaluating the parameters of interest has been statistically modeled and based on data collected on 30 Blonde d’Aquitaine heifers. The variables were obtained after a statistical pre-treatment (clustering of variables) to reduce the redundancy of the 62 initial variables. The sensitivity analysis evaluated the importance of each independent variable in the models, and a graphical approach completed the analysis of the relationships between the variables. Then, the models were used to generate virtual animals and study the relationships between the nutritional and organoleptic quality. No apparent link between the nutritional and organoleptic properties of meat (r = −0.17) was established, indicating that no important trade-off between these two qualities was needed. The 30 best and worst profiles were selected based on nutritional and organoleptic expectations set by a group of experts from the INRA (French National Institute for Agricultural Research) and Institut de l’Elevage (French Livestock Institute). The comparison between the two extreme profiles showed that heavier and fatter carcasses led to low nutritional and organoleptic quality.

La filière viande bovine française, qui est confrontée à de nombreux défis menaçant sa pérennité au sein des territoires, souhaite se relancer en proposant une montée en gamme de ses produits. Dans cette thèse, nous avons donc développé un outil d'aide à la décision pour accompagner les acteurs de la filière (transformateurs, distributeurs, etc.) dans la conception de nouvelles démarcations.Cet outil s'appuie sur une approche innovante pour gérer les antagonismes entre performance des bovins et la qualité de leur viande, et permet ainsi au décideur de choisir le compromis qui correspond le mieux à ses attentes.Cette recherche de compromis a été modélisée sous la forme d'un problème d'optimisation multi-objectifs qui est résolu grâce à la définition d'un ordre partiel entre les vecteurs de l'espace des objectifs : la dominance de Pareto.En l’absence de fonctions objectifs analytiques, des modèles de substitution ont été estimés à partir d’observations des objectifs pour des paramètres de production donnés. Pour assurer au décideur que la valeur optimale promise se réalisera à une occurrence contrôlée, une version de la Value-at-Risk adaptée à un problème multi-objectifs a été développée.Les nombreuses données nécessaires pour estimer les modèles de substitution sont issues d’expérimentations européennes sur les bovins. Celles-ci avaient étudié au minimum un aspect de la qualité de la viande issue du muscle Longissimus thoracis : la qualité sensorielle ou la qualité nutritionnelle, via le profil en acide gras du muscle.Une interface web libre d'accès a été conçue avec le package R Shiny afin de diffuser l’outil d'aide à la décision. Des cahiers des charges répondant à des enjeux que nous avons identifiés comme pertinents ont enfin été générés grâce à cette interface afin d’alimenter la discussion sur les potentielles applications de l’outil.