Using self-organizing maps for identification of roles in social networks

Abstract

In social networks the participants may be characterized by their roles. We understand roles as different patterns of link structure in the network. These roles describe the node and its activity in the network over time. Self-organizing maps (SOMs) - type of artificial neural-networks, are used for node's role identification and for discovery of all the roles present in the network. Different data preprocessing methods allow us to capture different aspects of roles. We show results of the experiment with a large scale co-authorship network constructed from a DBLP dataset.