Stability of local information-based centrality measurements under degree preserving randomizations

Abstract

Node centrality is one of the integral measures in network analysis with wide range of applications from socioeconomic to personalized recommendation. We argue that an effective centrality measure should undertake stability even under information loss or noise introduced in the network. With six local information-based centrality metric, we investigate the effect of varying assortativity while keeping degree distribution unchanged, using networks with scale free and exponential degree distribution. This model provides a novel scope to analyze the stability of centrality metric which can further find many applications in social science, biology, information science, community detection and so on.