Scanning network communities with power-law-distributed attributes

Abstract

Community detection has drawn significant attention as network generates big data every day. To simultaneously consider both attribute and structure cluster patterns, a scanning method [1] is recently developed to provide a statistical testing procedures. Some common distributions are considered in [1] except the power-law distribution, which network attributes are generally followed. This paper aims at extending the scanning method to be applied in a social network that its attributes follow power-law distribution. Besides the theoretical construction, an authorship network is used to demonstrate the proposed method.