Community Detection with Edge Augmentation in Criminal Networks

We study community detection in criminal networks and address the problem caused by intentionally hidden edges which hinder the performance of community detection. We make use of link prediction to demonstrate how the community structure of a network can be better identified by augmenting it with edges. We demonstrate the value of this method by showing this method delivers us better quality communities for real life drug trafficking networks. We discuss also the limitations of the approach, and importance of community detection for investigating of criminal networks.