An Evolutionary Algorithm for Network Clustering through Traffic Matrices

While network clustering is traditionally accomplished just relying on the topology of the network, the new traffic-aware clustering approach employs traffic matrices to take into account the intensity of the relationship between nodes. In the context of traffic-aware clustering we propose a new Evolutionary Clustering algorithm and compare it with the Spectral Filtering algorithm. We compare them using both the Modularity and the Traffic-aware Scaled Coverage metrics, and two real-world datasets, each made of 1000 traffic matrices, respectively from Abilene and Geant networks. Our experiments show that Evolutionary Clustering performs better on all traffic matrices, excepting a minor number of traffic matrices in the Abilene network when the Modularity metric is employed.