Better Ingredients, Better Pizza
One unusual application for social networks? Food.
That’s right — Lada Adamic, a computer scientist at the University of Michigan and at Facebook, has been working on network analysis of recipes, ingredients, cooking methods, and nutritional profiles. Her algorithm accomplishes something remarkable: it predicts with 80% accuracy the number of stars a recipe will receive on allrecipes.com using over 50,000 recipes and 2 million reviews. Using this information, she was able to construct a mapping of ingredients based on how often pairs of ingredients appeared together on a recipe.
The Food Network seems like a pretty aptly named channel now, doesn’t it? Adamic’s food network doesn’t just show us closely coupled ingredients, like milk and eggs, which in and of itself isn’t all the interesting; it can be used to predict which recipes will be successful, and even provide information about which food items can serve as substitutes for each other. This fascinating application of networks as predictive technologies really illustrates how anyone can learn to cook by appreciating the flexibility recipes can have. Novice cooks often fail because they don’t understand how ingredients are related to each other or are unfamiliar with what is proven to taste good. Or, they will find similar recipes on a website on allrecipes.com and be hesitant to choose one over the other. With a food network at the ready, they can get the hang of cooking techniques much faster and feel more comfortable experimenting.
Harvard physicists went a step further and analysed ingredient flavor profiles, which tell us a lot about prevalence, food categories, and chemical compounds within ingredients. The flavor profile network examines how different cuisines choose different categorical pairings, which are seen as strongly-connected large components in the graph. The interactions between these components provide instructions as to what works well together, and what doesn’t (or what hasn’t been tested).
Food networks are a more natural way to imagine they way we cook and eat, and demonstrate that network theory can be used to predict certain patterns across recipes and cuisines. The secret to better pizza is out!
Sources:
— cuckoo for coco puffs