ABSTRACT
G IVEN large social networks that have recorded millions of users and items (e.g., messages on Facebook/Twitter, articles on Shashdot/Digg),
billions of item adoption behaviors and millions of user connection behaviors, what is the underlying behavioral mechanism of the social network user? Can we accurately predict the most probable items that the user will adopt? Can we recommend to them what they really want? How can we alleviate
the challenging problem of high sparsity of the behavioral data? The general philosophy underlying these applications is a deep understanding of the multi-aspect nature of users’ behavioral intention.
