Correlation Between Researchers' Centrality and H-Index: A Case Study

Vincenza Carchiolo, Marco Grassia, Michele Malgeri, Giuseppe Mangioni

Abstract: The quality of research’ work is often very hard to assess, While controversial, the use of indices to measure the performance of the researchers and of institutions is widely accepted in scientific communities. In this paper, we build and study the co-authorship network of the MAT/05 “academic discipline” (as defined by Italian law), and investigate if the centrality measures from Network Science are enough to predict the h-index of a researcher, when fed to classic Machine Learning algorithms. Our results show that some models and combinations of features work better than others, and also that the partial network (e.g., without authors that do not belong to MAT/05) may not be enough to predict accurate h-index values.