Модели сети соавторства научного журнала. Часть 2

The paper analyzes two models of a scientific co-authorship network. The first model (𝑁𝑎) is built on binary relationships that arise between authors who have jointly created at least one article. This traditional model is presented in the form of an undirected graph, the vertices of which correspond to the authors, and the edges establish connections between them. The second model (𝑁𝑐𝑎), which takes into account group relations between the authors of a same article, is presented in the form of a bipartite graph. A higher-order architecture that generalizes pairwise interaction to the interaction of an arbitrary set of nodes helps expand the capabilities of modeling and understanding the behavior of a co-authorship system. This work continues the study and testing of methods for analyzing coauthorship networks [1, 2]. Methods for constructing both models based on data extracted from an XML archive of scientific journal articles are presented. The basic parameters of the models were measured, for results see table 2. It should be noted that network contains significantly fewer edges than 𝑁𝑎, which is important when studying large amounts of data. It is revealed that the distribution of degrees of the vertices of both subsets of bipartite representation corresponds to the power law, see fig. 2. The average distance between nodes is approximately six, i. e. we can apply the term “small world” [40] to it. The centralities of the actors [32-36] that determine their potential to influence processes occurring in the network are estimated. Using the small component as an example (fig. 3), it is shown that the nodes are more differentiated within the corresponding measure, see tables 3-4. Сo-authorship patterns are identified based on a study of bicliques in a bipartite representation. Analysis of bicliques showed that for this co-authorship system, the most common pattern is the presence of pairs of authors who jointly published two articles. The next most common pairs are the more stable pairs, who published three articles each, and the triplets of authors, who published two articles each. Fig. 4 illustrates the distribution of 𝐾𝑖,2. Further examination of bicliques can determine which authors participate in multiple groups and whether the group of authors is expanding over time.