Using diet optimization and machine learning for the design of healthy and acceptable menu plans

The success of dietary plans relies on understanding and modelling consumer acceptance, yet quantifying this poses a challenge due to the complexity of individual preferences. Recent research is focused on deriving acceptability constraints directly from data, as demonstrated by its application in designing food baskets with a limited number of commodities. In this study, we applied diet optimization with machine learning to the more complex task of menu planning. This involved considering hundreds of potential food alternatives and assessing their compatibility within a meal using a recipe completion algorithm. Compared to the traditional diet modelling approach of food group filtering, the recipe completion model delivered diets with either higher nutritional adequacy or greater substitute acceptability, depending on the number of food groups used in the traditional method. While more research is needed to further improve the acceptability of substitutions, combining diet optimization with recipe completion presents a promising approach to enhance the nutritional adequacy of individual diets while maintaining the acceptability of food combinations within meals. • New method that combines diet modelling with recipe completion to model acceptability. • Food items are substituted considering meal context. • Acceptability of food substitution, applied to real-world data, is evaluated.