Performance Analysis of an Algorithm for Computation of Betweenness Centrality

Abstract

AbstractIn social network analysis, graph-theoretic perceptions are used to realize and explain social experience. Centrality indices are essential in the analysis of social networks, but are costly to compute. An efficient algorithm for the computation of betweenness centrality is given by Brandes that has time complexity O(nm + n2logn) and O(n + m) space complexity, where n, m are the number of vertices and edges in a graph, respectively. Some social network graphs are invariably huge and dense. Moreover, size of memory is rapidly increasing and the cost of memory is decreasing day by day. Under these circumstances, we investigate how the computation of centrality measures can be done efficiently when space is not very significant. In this paper, we introduce a time efficient and scalable algorithm for the accurate computation of betweenness centrality. We have made a thorough analysis of our algorithm vis-à-vis Brandes’ algorithm. Experimental results show that our algorithm has a better performance with respect to time, but at the expense of using higher memory. Further performance improvement of our algorithm has been achieved by implementing it on parallel architectures.KeywordsSocial networksbetweenness centrality