Detection And Application Of Research Preference Community Based On Author-Topic Network

Under the context of the country put a high value on science on the research activities and scholars,and try to exploit fundamental research supporting science and technology,it is meaningful to analyze research activities from scientific papers.Community detection provide a new perspective on exploring research activities.Finding the community of scholar network can discover law of the subject and scholars and apply to scholar recommendation.Regarding the community detection of scholar networks,scholars' researches cover basic community detection to community evolution,as well as algorithm optimization and application expansion for the characteristics of scholar networks.Due to discover academic interest groups and promote cooperation between scholars,Initially,the thesis takes the relationship of scholars' common research interests as object and take the bibliography of library and information science as sample.The thesis focuses on two aspects:(1)how to represent to research preference of scholars and to find the community in the author-topic network.The bibliographic abstract is trained through the author's topic model,and the probability distribution of the author on the topic is obtained,which is used to represent the author's research interest.Moreover,using the LPAwb+ method to detect community structure in the network.The community detected by LPAwb+ has an advantage,which is help to explain why scholars in this community.(2)Explore the application of potential collaborator recommendations based on research preference communities.Introduce the perspective of cooperation to explore the influence mechanism of research preference communities,and use ERGM to quantitatively analyze the influence of research preference communities to cooperative networks.And the influence of the research preference community on the formation of the cooperative network is integrated into the link prediction,as the community weight of the public neighbors.The thesis verifies that the attributes of the research preference community really have a positive effect on cooperation network.Scholars in the same research preference community are more likely to cooperate,and there are differences in the tendency of different research preference communities.Use the character in potential collaborator recommendations will increase the rationality.The lack of research lies in the fact that the ATM does not include a time dimension,thus the thesis is short of the discussion on the evolution of the research preference community and the development trends of library and information science.