A density based algorithm for community detection in hyper-networks

Abstract

We propose an efficient community detection algorithm for networks that comprise more than one entities, such as users, tags and items, with ternary or higher relations between them. Such networks are also known as multi-partite and can be used for representing social tagging systems but also the activity in streaming media. Detecting communities in multi-paritite networks entails different challenges than in simple networks. The proposed algorithm is able to detect crisp or overlapping communities, and is applied on four data sets from social tagging systems and Twitter, and is compared with other multi-partite community detection algorithms.