Triad-Based Comparison and Signatures of Directed Networks

Abstract

We introduce two methods for comparing directed networks based on triad counts, called TriadEuclid and TriadEMD. TriadEuclid clusters the Euclidean distance between triad counts, whereas TriadEMD is an adaptation of NetEMD for directed networks. We apply both methods to cluster synthetic networks, a set of web networks including google, twitter, peer-to-peer, amazon, slashdot and citation networks, as well as world trade networks from 1962-2000. Furthermore, we find signature triads and signature orbits for each type of networks in our data, which show the main triad and orbit contributions of the networks when comparing them to the other networks in the respective data set.