Community detection in social networks with an application to COVID-19 data

Abstract

Social network analysis (SNA) is a data analytic field that investigates hidden structures using the baseline of networks and graph theory. It helps to understand the nature of creating connections between the objects. Within a network, there can be multiple sub-networks which are called as ‘communities’, and there are various algorithms to find communities within a network. In this thesis, we analyze an epidemic spread using social network analysis, based on the data from the COVID-19 outbreak across the world and in Canada. We assess the nature of the spread of this virus by detecting communities using different community detection methods which can be applied on directed networks; Louvain, Label propagation, Infomap, and Spinglass algorithms. We then evaluate the performance of the community detection algorithms using simulation studies. We also assess the impact of the density and sparsity of the network on community detection by introducing a novel random partition graph generator using a mixture of two Gaussian distributions.