Friends are forever? Evolution of active friendship clusters in online social networks

The goal of this paper is to study how friendship clusters evolve in online social networks. Results obtained from our work on Facebook data indicate that the set of friends who actively interact during a particular time interval is only a fraction of the total number of listed friends and this set of active friends tends to evolve with time. Interaction footprints on the timeline are used to cluster such active friends. Since there is a temporal variation in these clusters, it is modeled as a graph, wherein nodes represent cluster memberships and edge weights represent the number of common members between a pair of nodes. Novel methodologies for predicting the size and membership of the next active cluster based on the features extracted from past interaction of a user are introduced and their performance evaluated.