Web Of Science-based Research On Author Cooperative Network Analysis

With the continuous development of Internet technology and the continuous emergence of academic achievements,scholars can use various analytical methods to explore hidden relationships between academic achievements and explore the laws of academic cooperation.Among them,the analysis of academic cooperative relations has gradually become a research hotspot.Research on the author cooperative network can understand the reasons for the formation of author cooperation and the development trend in the research field,help scholars to find partners more conveniently,and promote the formation of scientific research cooperation.In this paper,the laws and influencing factors of scholars' cooperation are first analyzed.On this basis,Methods based on link prediction and network-based representation learning are used to establish the scholars' cooperative relationship prediction model,and the scholars' external attributes are added to the traditional model.These new features have been experimentally proved to be effective in improving the prediction accuracy of the model.This paper summarizes the previous literature,based on statistical methods,selects two types of influencing factors of the author's cooperative relationship including cooperative network topology characteristics and author characteristics.Cooperative network topology features select link prediction features and network representation learning features.The link prediction features include Common Neighbours(CN),Adamic-Adar(AA),Resource Allocation(RA),Jaccard,Katz,Graph Distance(GD),Sim Rank;the network representation learning feature is the author vectors calculated by the LINE algorithm.The author's characteristics select the author's academic age,the number of articles published,the cooperation rate,research interest,institution and region.This paper uses the Web of Science core database library and information field(1945-2018)data set to calculate the link prediction features,network representation learning features,and author's external features.Two models are established based on link prediction and network representation learning.The author's external features are introduced into the above two models to form four prediction models.Through the experiments of the four models,based on the Precision index,it is found that the model prediction effect is the best when the link prediction feature and the author's external feature are used,and the accuracy can be achieved.By comparing the Gini coefficient of each author's features output by the model,it is found that the author's cooperation rate difference,the number of authors' papers and the academic age difference are negatively correlated with the author's cooperation.There is a positive correlation between the author's cooperation or not.