Deliverable D 7.4: Toward modelling SHARP diets, based on nutritional adequacy, sustainability metrics and population diversity parameters
Diet modelling has been dominated by linear programming models for many years, however their success has been limited, while their inability to extract value from data in our information-driven world has become readily apparent. Increasing consumers’ diet healthiness has been the primary task of almost all diet models, however to actually change patterns of consumers’ purchasing behavior, models have to learn their preferences, so as to recommend diet alternatives that are both healthy, and appealing. We present our data-driven approach that leverages food item similarities as the main building blocks of diet recommendations, which arguably represents a paradigm shift in the way we optimize diets. Furthermore, we switch from exploiting what is known to be preferable (i.e. what we observe in a consumer’s diet), to exploring what is likely preferable, thereby allowing our diet model to “think outside the box”.
Take a look