User retention is important to the success of online social media, particularly in virtual world settings where users shape one another’s online experience. We study a rich set of variables, including social network and group membership, chatting, and transactions, in order to predict which users will stay and which ones will leave. We find that simple variables directly measuring the intensity and diversity of a user’s interaction with others are most predictive.
ChunYuen Teng, Lada A. Adamic