Academic Network Visualization System Based On Relational Mining

With the development of computer science and the popularity of the Internet,research activities have gradually become more complex,diversified and interdisciplinary.Academic collaboration has gradually become the main working method for scientific research workers to carry out academic research and solve scientific problems.As a special social network,academic network attracts wide attention with its characteristics of nonlinearity,complexity,diversity and mass.For academic networks,the common forms of representation of data such as tables,texts,etc.,have made it difficult for people to understand the information contained in academic networks clearly and intuitively.Therefore,how to achieve the visualization of academic networks effectively,achieve the visualization of internal structure of academic networks visually,and mine valuable information hidden inside the academic networks is of great significance for academic networks research.Meanwhile,academic influence is a hot topic in academic network research and an important part of academic network visualization.As the significant force to promote scientific research progress,in addition to scholars' own scientific research output(papers,books,patents,etc.),their students are also an important part of their contribution.Different from the scholars' influence evaluation research based on the scholar's own scientific research output,this paper innovatively evaluates the influence of scholars from the perspective of the mentor-student.This research has important significance for hot topics such as mentor recommendation and scholar influence assessment.Firstly,based on Microsoft Academic Graph(MAG),this paper visualizes the scholar's relationship networks in the form of graph by analyzing the scholar's relationship data.For the complex relationships in academic networks,based on the full reading of literature,research and existing related work,this paper divides the relationship in academic network into three categories: citation relationship,collaboration relationship and advisor-advisee relationship.In the visualization display,the three relationships are displayed in the form of cluster network graph,pie graph,tree graph,pie graph and colony map.Then,in the process of visualizing the advisor-advisee relationship,a scholarly influence evaluation method based on the mentor-student dichotomy network was creatively proposed.By processing the Microsoft Academic Graph dataset and the Web of Science journal impact factor dataset,Several characteristics of students are extracted as evaluation indicators.The comprehensive score of students is obtained by using the comprehensive evaluation method combined with the weighted academic age of students.Next,establish a mentor-student dichotomy network,use the student's comprehensive score as the weight of the side,calculate the influence of the instructor by random walk,and then combine the mentor's academic age to obtain the final influence index of the mentor.Finaly,display it in the system as a ranking manner.The visualization part of this paper and the student-based scholars' impact assessment have been deployed as functional modules in the academic network visualization system.