Connecting retailer to recipes data: a network analysis of food consumption in London

This research offers an insight into food networks analysis, focusing on the gap between digital representation of food consumption and actual consumption with a data-driven investigation. Data come from a large British retailer and contain the record of all the products bought in the area of Greater London during a period of one year. By extracting the ingredients from the purchased products, a number of networks of ingredient co-occurrence are built and backbone algorithms are explored. The goal would be to check whether food networks from literature, which are related to ingredient co-occurrence in recipes, share any broad-strokes features with the co-occurrence in products, or whether the networks fundamentally distinct. This is done via staple ingredients analysis as long as usual metric analysis such as degree distribution, distances, communities. The findings reveal that the products-based network presents more cross-cultural paths with respect to recipes-based network that shows instead a strong dichotomy between western and eastern cuisines. Future research is to test if properties of the network constructed via retailer data, representation of actual food consumption, are connected with health indicators and, starting from this, how healthy diets can be recommended.