H.G.J. Gremmen

Consumer-Oriented New Product Development in Fruit Flavour Breeding - A Bayesian Approach

Taking consumer quality perceptions into account is very important for new-fruit product development in today’s competitive food market. To this end, consumer-oriented quality improvement models like the quality guidance model (QGM) have been proposed. Implementing such models in the agro industry is challenging. We propose the use of Bayesian structure equation modeling (SEM) for parameterizing the quality guidance model, allowing for the integration of elicited expert knowledge. Such casual modeling would furnish important insights for determining the optimal fruit product in terms of consumer flavor-quality perceptions. In the context of tomato breeding, where we have data about metabolites, sensory-panel judgments, and consumer flavor-quality perceptions, we estimated a benchmark Bayesian SEM using non-informative priors, starting from an initial causal model derived from the data with a score-based Bayesian network (BN) learning algorithm. The results so far have given some indication of the importance of accounting for consumer heterogeneity in the modeling process.