"Clumpiness" Mixing in Complex Networks

TL;DR

This paper introduces new measures of network clumpiness that account for the distribution of central nodes, providing a more comprehensive classification of complex networks than previous measures.

Contribution

The paper proposes a novel clumpiness coefficient that considers separated central nodes and demonstrates its effectiveness in classifying real-world networks.

Findings

Successfully classifies 30 real-world networks into four categories.

Differentiates between Erdos-Renyi and Barabasi-Albert models.

Provides bounds and relationships for the new clumpiness measures.

Abstract

Three measures of clumpiness of complex networks are introduced. The measures quantify how most central nodes of a network are clumped together. The assortativity coefficient defined in a previous study measures a similar characteristic, but accounts only for the clumpiness of the central nodes that are directly connected to each other. The clumpiness coefficient defined in the present paper also takes into account the cases where central nodes are separated by a few links. The definition is based on the node degrees and the distances between pairs of nodes. The clumpiness coefficient together with the assortativity coefficient can define four classes of network. Numerical calculations demonstrate that the classification scheme successfully categorizes 30 real-world networks into the four classes: clumped assortative, clumped disassortative, loose assortative and loose disassortative networks. The clumpiness coefficient also differentiates the Erdos-Renyi model from the Barabasi-Albert model, which the assortativity coefficient could not differentiate. In addition, the bounds of the clumpiness coefficient as well as the relationships between the three measures of clumpiness are discussed.