A Topological Data Evolution Based Method to Predict Links in Social Networks

Abstract

Link prediction is a graph mining task that aims to identify pairs of non-connected vertices that have a high probability to connect in the future. This task has been frequently implemented by recommendation systems that suggest new interactions between users in social networks. In general, the state-of-the-art link prediction methods only consider data from the most complete and recent state of the network. They do not take into account information about the existing topology when new edges were added to the network's structure. This study raises the hypothesis that recovering such data may contribute to building predictive models more precise than the available ones since those data enrich the description of the application's context with examples that represent exactly the kind of event to be foreseen: the appearance of new connections. Hence, this paper evaluates such hypothesis. For this purpose, it proposes a link prediction method that is based on the historical evolution of the topologies of social networks. Results from experiments with ten real coauthorship social networks reveal the adequacy of the proposed method and the confirmation of the raised hypothesis.