Network Classification in Temporal Networks Using Motifs

  • 2018-07-10 16:09:29
  • Kun Tu, Jian Li, Don Towsley, Dave Braines, Liam D. Turner
  • 2


Network classification has a variety of applications, such as detectingcommunities within networks and finding similarities between those representingdifferent aspects of the real world. However, most existing work in this areafocus on examining static undirected networks without considering directededges or temporality. In this paper, we propose a new methodology that utilizesfeature representation for network classification based on the temporal motifdistribution of the network and a null model for comparing against randomgraphs. Experimental results show that our method improves accuracy by up$10\%$ compared to the state-of-the-art embedding method in networkclassification, for tasks such as classifying network type, identifyingcommunities in email exchange network, and identifying users given theirapp-switching behaviors.


Introduction (beta)



Conclusion (beta)