Spilling the beans: Food recipe popularity predictionusing ingredient networks

Food plays a central role in all of our lives, it affects our health and even our mood. There are millions of different (online) recipes to choose from, a lot of which haven't been vetted yet. This research aims to glean new insights in which features drive the popularity of recipes by way of network analysis, and use these insights to train a predictive model. While the best of a pair of similar recipes can be determined with an accuracy of 90%, a more general rating predicting proves to be a much tougher nut to crack. We haven't been able to accurately predict a general rating for recipes, but we believe we can provide some food for thought.