面向科技大数据的科研团队精准立体画像生成方法

Accurate Stereo Portrait Generation Method of Scientific Research Team for Scientific and Technological Big Data

针对科研团队精准立体画像构建的现实需求,本文提出了面向科技大数据的科研团队精准立体画像生成方法。该方法包含科研团队识别、科研团队研究主题预测、科研团队精准立体画像生成三个模块。基于迭代的中间中心度排名方法识别科研团队负责人,通过2-派系和滚雪球方法识别科研团队成员,实现科研团队的识别。综合考虑研究领域内科研团队研究主题间的交互影响,提出基于图注意力网络的主题融合的科研团队研究主题预测方法,通过建模相同研究领域的团队间的主题关联,提高科研团队的研究主题预测准确率。科研团队精准立体画像生成模块通过科研团队简介生成、科研合作关系构建、科研团队研究主题及主题趋势词云生成科研团队的精准立体画像。基于提出的方法在通信领域的科研成果数据集上识别出科研团队,预测科研团队的研究主题,生成科研团队精准立体画像并进行可视化展示,验证了所提方法的有效性。

In view of the practical demand of accurate stereo portrait construction of scientific research team,this paper proposes an accurate stereo portrait generation algorithm of scientific research team.The algorithm includes three modules:research team identification,research topic prediction and research team portrait generation.The leader of the scientific research team is identified based on the iterative middle centrality ranking method,and the members of the scientific research team are identified through the 2-faction and snowball methods,so as to realize the identification of the scientific research team.Considering the interaction between the research topics of the research team in the research field,a topic prediction method based on the topic fusion of the graph attention network is proposed to improve precision of topic prediction.The research team portrait generation module generates the accurate stereo portrait of the research team through the generation of the research team profile,the construction of research cooperation relationship,and the construction of the research team topic and trend cloud.Scientific research team is identified on the data set of scientific research achievements,and the accurate stereo portrait of scientific research team is generated and visualized.Experiments verify the effectiveness of the proposed method.