Correlation Analysis of Local Graph Metrics

Investigating the correlations between graph metrics in various networks is a prosperous research topic that only starts to evolve. Knowledge about graph metrics interrelations may be beneficial in many other aspects of networks studying, starting from optimizing computations to even detecting particular network types by looking at their correlation maps. In this article, we analyze interrelations between a wide range of local network metrics, both on simulated and on real data, attempting to find and explain differences in those relations. Several crucial discrepancies between different graph generating models were found, which could be a consequence of specific underlying graph-generating mechanisms. Moreover, as opposed to modeled networks, where metrics were correlated to a large extent, real-world graphs demonstrated less intensive interdependencies between metrics. Nevertheless, the majority of strong correlations found in real data coincide by sign with those found in modeled data. Thus, differences between graph types are likely to be accumulated in just several metrics, which would be helpful in the task of determining network type by its correlation matrices