Diversity Index of Academic Community Ecosystem by Co-authorship Analysis with Granger Causality

Abstract

Compared to individual-level analysis, communitylevel analysis provides a new perspective to inspect network structure. It focuses on modeling the evolving relationships between communities. Intuitively, community-level analysis is a generalization of individual-level analysis. It reflects a macroscopic evolution of a network and reduces the overfitting of individual analysis to some degree. In this paper, we investigate the co-authorship characteristics between different affiliations in academic social networks and then adopt the weighted multigraph model to establish the coauthorships between communities. Subsequently, we define the Co-authorship Factor (CF) for each pair of communities and then propose the modified Shannon Co-authorship Diversity Index (SCDI) and Renyi Coauthorship Diversity Index (RCDI) to measure the diversity of co-authorship ecosystem of a certain community. Finally, we apply the Granger causality to model the mutual co-authorship influences between communities along time. We verify our proposed indexes on real dataset which is mainly based on the DBLP and Microsoft Academic Graph (MAG) datasets. Keywords—Community analysis; Granger causality; Academic social