Global and local community memberships for estimating spreading capability of nodes in social networks

Simon Krukowski,Tobias Hecking

doi:10.1007/s41109-021-00421-3

Abstract

The analysis of spreading processes within complex networks can offer many important insights for the application in contexts such as epidemics, information dissemination or rumours. Particularly, structural factors of the network which either contribute or hinder the spreading are of interest, as they can be used to control or predict such processes. In social networks, the community structure is especially relevant, as actors usually participate in different densely connected social groups which emerge from various contexts, potentially allowing them to inject the spreading process into many different communities quickly. This paper extends our recent findings on the community membership of nodes and how it can be used to predict their individual spreading capability (Krukowski and Hecking, in: Benito, Cherifi, Cherifi, Moro, Rocha, Sales-Pardo (eds) Complex networks & their applications IX. Springer, Cham, pp 408–419, 2021) by further evaluating it on additional networks (both real-world networks and artificially generated networks), while additionally introducing a new local measure to identify influential spreaders that—in contrast to most other measures, does not rely on knowledge of the global network structure. The results confirm our recent findings, showing that the community membership of nodes can be used as a predictor for their spreading capability, while also showing that especially the local measure proves to be a good predictor, effectively outperforming the global measure in many cases. The results are discussed with regard to real-world use cases, where knowledge of the global structure is often not given, yet a prediction regarding the spreading capability highly desired (e.g., contact-tracing apps).

Highlights

The study of spreading processes on networks has a long research history in information diffusion in social networks (Guille et al 2013), computer communication (Balthrop et al 2004), and epidemiology (Nowzari et al 2016)
This paper extends our recent findings on the community membership of nodes and how it can be used to predict their individual spreading capability
We approach two goals: Firstly, extend our understanding of our recently introduced measure of community centrality, and secondly, develop a new measure to predict the efficiency of the spreading process which can be used without knowledge of the global network structure

Summary

Introduction

The study of spreading processes on networks has a long research history in information diffusion in social networks (Guille et al 2013), computer communication (Balthrop et al 2004), and epidemiology (Nowzari et al 2016). Being referred to as immunisation or attack strategies, they usually examine which nodes exert the most influence on the information diffusion process, i.e., are the most influential spreaders, in order to either immunise or attack them and better control the spreading (Cherifi et al 2019; Magelinski et al 2021) In this context, topological properties of nodes (e.g., centrality, community membership) are of interest. A important topological feature which influences the efficiency of the spreading process is the community structure of networks and the resulting properties of nodes (Rajeh et al 2021; Ghalmane et al 2019a, b; Kitsak et al 2010) To this end, Kitsak et al (2010) showed that the most efficient spreaders within a network are not necessarily the most central nodes (i.e. nodes with the highest degree), but the ones located in densely connected cores of the network indicated by a high k-shell index. As such, considering the community structure can improve the prediction of the spreading process and the immunisation of relevant nodes (Peng et al 2020; Ghalmane et al 2019a)

Objectives

Methods

Conclusion