Network of Culinary Practice
With two of my group members, I manipulated network data about different chemical pairings of ingredients used across the globe to capture the characteristics of culinary practice. Inspiration for this project came from the report “Flavor Network and the Principles of Food Pairing” in which the researchers Yong-Yeol Ahn, Sebastian E. Ahnert, James P. Bagrow and Albert-László Barabási present a network science approach to understanding culinary practices. Even though there were several moments where our research did not go as planned due to our lack or knowledge and skills on network analysis, I believe this project still holds a powerful statement that challenges the limit of data science on culinary practice that tend to be perceived as solely subjective field.
Network visualization is something we struggled the most. We used Louvain Community Detection method, an algorithm that’s said to be the best for the data set of this size. The figure to the right is the back-bone version of the left, meaning that all redundant ingredients are removed. The colors are categorized as groups of ingredients that shares the flavor compounds the most with.
This graph organizes all the ingredients in unweighted degree (a measurement that shows how many other ingredients that specific ingredient share flavor compounds with). The ingredient with the highest degree is black tea, sharing flavor compounds with 989 ingredients.
