Структура и параметры невзвешенной сети соавторства на основе данных БД REPEC

The structure and parameters of the unweighted co-authorship network based on DB REPEC data

In this paper we investigate the structure of scientific collaboration based on co-authorship in publications retrieved from the RcPEc database. The main attention is paid to the co-authorship network: nodes correspond to authors, and two authors arc considered connected if they arc coauthors of at least one publication. The network is represented by the undirected unweighted collaboration graph. The co-authorship in articles provides a window on patterns of collaboration within the academic community. The study of scientific collaboration networks is carried out in two main directions. Empirical measurements provide detailed characteristics of the network: statistical properties of the distribution of degrees of nodes, global network parameters, nodes centralities (Newman 2001a; 2001b; 200lc). The study of the dynamic properties of real networks and network models makes it possible to identify the structural principles that govern the evolution of networks; dynamic properties, in turn, can explain static ones (Barabasi et al., 2001; Savic et al., 2017). Co-authorship networks arc used to identify research groups and the most important researchers and to predict their scientific success; to classify journals by degree of co-authorship and to make maps of science. Co-authorship networks can be constructed for different components of analysis, such as researchers, institutions, and countries. We use a researcher as the unit of analysis. Let P be the set of publications under consideration and V - the set of the authors of these publications. We assume that each publication in P has at least two authors. For i,j e V let us define the binary relation of co-authorship Rca: i Rcaj = (3p e P) the author i is one of the authors of p and the author j is one of the authors of p. The co-authorship network is the weighted undirected network Nca = (V, E,w), V - the set of the authors, E C V x V - the set of weighted edges, e = (i, j) e E, if iRcaj. In the simplest case, w(e) = 1, regardless of how often these authors are coauthors and how much each contributed to coauthored publications, that is, the network can be considered unweighted. Then the graph corresponding to the unweighted network is represented bv the (0,1) adjacency matrix U = (uij). The articles used in this study were retrieved from the RcPEc database (REPEC). In order to identify authors uniquely and to infer actual author identity we use the author „ profile“ that authors create basing on the Author Service provided by the RcPEc database (similar to Google Scholar). So we have |P| = 91113 publications with more than one author and |V| = 32 434 authors of these publications. In total, these authors have published 364 979 articles. The publications refer to the period from 1954 to 2019. We presented the base statistics such as papers per author, authors per paper, maximal authors per paper and collaboration indices Cl, DC, CC, MCC. We also measured the network parameters such as the size of the largest component, network density, average distance, maximal distance and clustering coefficient. It was found that the distribution of numbers of coauthors for authors follows power law with exponent 7 = 1.3. The study showed that in the collection of publications under consideration the fraction of coauthored publications is small (25%) and the prevailing trend is the presence of two coauthors in a publication (77%). The most authors are indirectly connected to each other - the maximum component includes 90 % of authors. The co-authorship network under consideration is scale-free and shows the „small world“ effect.