Eigenvector Centrality Measure Based on Node Similarity for Multilayer and Temporal Networks

Laishui Lv,Jiahui Zhang,Wei Xue,Kun Zhang,Xun Li,Ting Zhang

doi:10.1109/access.2019.2936217

Laishui Lv, Jiahui Zhang + Show 4 more

Open Access

https://doi.org/10.1109/access.2019.2936217

Copy DOI

Abstract

Centrality of nodes is very useful for understanding the behavior of systems and has recently attracted plenty of attention from researchers. In this paper, we propose a new eigenvector centrality based on node similarity for ranking nodes in multilayer and temporal networks under the framework of tensor computation, referred to as the ECMSim. We define a fourth-order tensor to represent the multilayer and temporal networks. The relationships between different layers(or time stamps) can be depicted by using node similarity. Based on the defined tensor, we establish the tensor equation to obtain nodes centrality values. The nodes centrality values also can be viewed as the Perron eigenvector of a multi-homogeneous map. Furthermore, we show the existence and uniqueness of the proposed centrality measure by existing results. Numerical experiments are carried out to demonstrate that the proposed centrality outperforms some existing ranking methods.

Highlights

Evaluation the importance of nodes is a core part of studying the network topological structure and has a wide range of applications in many fields, such as identifying novel drug targets in biological systems [1], extracting communities in social networks [2] and ranking the results of search engines [3]
In this paper, motivated by the idea of Taylor [28] and Yin [29], we propose a novel eigenvector centrality based on the node similarity for multilayer and temporal networks by introducing a fourth-order tensor to represent multilayer and temporal networks, referred to as the ECMSim
TENSOR REPRESENTATION OF MULTILAYER AND TEMPORAL NETWORKS In this paper, we focus on two types of complex networks: multilayer network, in which nodes connected by different types of links; and temporal networks, in which nodes and edges appear and disappear over time

Summary

INTRODUCTION

Evaluation the importance of nodes is a core part of studying the network topological structure and has a wide range of applications in many fields, such as identifying novel drug targets in biological systems [1], extracting communities in social networks [2] and ranking the results of search engines [3]. In [28], Taylor et al think that the connection between the nodes is dependent on the information about neighbour time layers, and relies on the other time layers Based on this consideration, a new supra-centrality matrix is proposed to represent multiplex and temporal networks. In this paper, motivated by the idea of Taylor [28] and Yin [29], we propose a novel eigenvector centrality based on the node similarity for multilayer and temporal networks by introducing a fourth-order tensor to represent multilayer and temporal networks, referred to as the ECMSim. For multilayer networks, we consider that there exist inter-layer edges between a node in one layer and that same nodes in other layers.

TENSOR REPRESENTATION OF MULTILAYER AND TEMPORAL NETWORKS

THE PROPOSED CENTRALITY

2: Output

NUMERICAL EXPERIMENTS

WORKPLACE TEMPORAL NETWORK

CONCLUSION