Chain of Influencers: Multipartite Intra-community Ranking

Abstract

Ranking of vertices is an important part of social network analysis. However, thanks to the enormous growth of real-world networks, the global ranking of vertices on a large scale does not provide easily comparable results. On the other hand, the ranking can provide clear results on a local scale and also in heterogeneous networks where we need to work with vertices of different types. In this paper, we present a method of ranking objects in a community which is closely related to the analysis of heterogeneous information networks. Our method assumes that the community is a set of several groups of objects of different types where each group, so-called object pool, contains objects of the same type. These community object pools can be connected and ordered to the chain of influencers, and ranking can be applied to this structure. Based on the chain of influencers, the heterogeneous network can be converted to a multipartite graph. In our approach, we show how to rank vertices of the community using the mutual influence of community object pools. In our experiments, we worked with a computer science research community. Objects of this domain contain authors, papers (articles), topics (keywords), and years of publications.