Recomender System Based on the Publication Graph Scientific and Pedagogical NRU HSE Staff

Co-authorship network is an undirected graph which nodes represent authors and edges represent collaborations between authors. The paper is devoted to building a recommender system that is based on a co-authorship network that was obtained from a database of all publications written by NRU HSE authors. Nodes’ and edges’ attributes were extended with data that was received from Scopus SJR Ranking database and a database of NRU HSE staff. We can analyze quality of papers and figure out some dependencies using the data. The goal of the paper is to build the recommender system and analyze the co-authorship network. The major problems of the research are data preprocessing, co-authorship network building, computation of basic metrics of the network and their interpretation, building and evaluation of the recommender system. As a result, of the network analysis some patterns of publication activity and quality of publications were figured out. More specifically, some departments that have authors that did not publish enough papers to fit recommendations of science committee of NRU HSE were detected. Moreover, graphs of departments’ areas interaction were plotted. Strong collaborations with technical studies departments for writing papers that are indexed in Scopus were emphasized on the graphs. Finally, the recommender system for predicting collaborators among staff of the university was built and showed good performance on the co-authorship network.