Comparative Study of Clustering Approaches in Rumor Source Localization Algorithm in Social Networks

Social network usually consists of a set of people acting as nodes in which one node is connected to other nodes internally. These network nodes can be visualized and obtain a mathematical model of these relationships (edges) by analysing the social network. Large social networks can be divided into different communities which are tied together with strong and weak edges. This immensely populated and dense social network, however, is vulnerable to rumor dissemination among communities in the network. To detect the source of information, a two stage algorithm has been proposed by previous studies. This algorithm relies on clustering for detecting communities in the network. The main aim to improve this algorithm is by comparing the results of different community detection algorithms and executing the algorithm on a variety of datasets. The previously developed two-staged algorithm uses the Louvain clustering for this purpose. It has been proposed to replace the existing approach of clustering with a better performing algorithm. The results of these algorithms are demonstrated by computing various metrics such as average distance error (hop error) and the probability of correctly localizing the source. Finally, it has been demonstrated that the replacing Louvain approach with Girvan-Newman Algorithm in the source estimation algorithm improves its performance in terms of the accuracy of correctly estimating the source of rumor in a large-scale network.