Power-law distributions of attributes in community detection

Abstract

Community detection has drawn significant attention as new media generates big data every day. To provide statistical testing procedures for community detection in social networks, a scanning method has been developed based on the likelihood of Poisson random graph. However, the scan statistics did not consider detecting communities of the attributes with power-law distribution. Power-law distribution, generally followed by network attributes, is conspicuous in many scientific situations. This paper aims at extending the scanning method to analyze a social network in which attributes follow power-law distribution. Besides the theoretical construction, simulation studies are performed to verify the feasibility of the proposed method, and an authorship network is used to demonstrate the proposed method.