Investigating and predicting online food recipe upload behavior

Abstract

Studying online food behavior has recently become an active field of research. While there is a growing body of studies that investigate, for example, how online recipes are consumed, in the form of views or ratings, little effort has been devoted yet to understand how they are created. In order to contribute to this lack of knowledge in the area, we present in this paper the results of a large-scale study of nearly 200k users posting over 400k recipes in the online recipe platform Kochbar.de. The main objective of this study is (i) to reveal to what extent recipe upload patterns can be explained by socio-demographic features and (ii) to what extent they can be predicted. To do so, we investigate the utility of several features such as user history, social connections of the users, temporal aspects as well as geographic embedding of the users. Statistical analysis confirms that recipe uploads can be explained by socio-demographic features. Extensive simulations show that among all features investigated, the social signal, in the form of friendship connections to other users, appears to be the strongest one and henceforth is the best to predict what type of recipe will be uploaded and what ingredients will be used in the future. The research conducted in this work contributes to a better understanding in online food behavior and is relevant for researchers working on online social information systems and engineers interested in predictive modeling and recommender systems.