Discovering Latent Behavioral Roles in Dynamic Social Networks

A new method to identify latent behavioral roles in dynamic social networks is proposed. Instead of assuming homogeneous behavior of all actors in an event-based social networks data set, the new approach analyzes the individual choice patterns of actors and clusters them based on their behavioral similarity. The new method is applied to a cell phone communication network. In this data set, three significantly different behavioral roles are identified. The choice patterns and the relations between actors of different roles are explored and discussed.