Mining Of Academic Socielty Based On Multi-Relationship Graph

With the continuous advancement and development of the informationization process,the digitalization and networking of today’s society are becoming more and more popular.Problems such as the massive accumulation of data and the high data dimension accompanying the era of big data have also brought problems to information acquisition in various fields.At present,the academic circle is generally divided into small academic groups by the unit,teacher relationship or academic activities.Scholars outside the group who want to understand the research results of other groups in the field often need to spend a lot of energy.Which has affected the sharing of academic resources,inheritance and exchanges between researchers to a certain extent.Therefore,an efficient academic circle mining method is needed to accurately describe the degree of association among researchers,promote academic exchanges among researchers,and provide reference for academic activities for young researchers.A traditional tool for mining academic circles is the citation relationship.There are several relationships between papers: direct-indirect reference,coupling,co-authoring,and citing.Through these relationships,a paper citation network can be constructed.The analysis of the degree of association between nodes in the network can be completed by using the quantitative indicators of the above relationships,so as to obtain a paper citation relation network with quantitative representation of the correlation strength between the paper nodes.Aggregating on the paper network can get a network of citation relationships about the author.The author’s citation relationship network represents the citation relationship between authors.On the other hand,the participation of researchers in the same academic conference and co-authorship is a natural academic community attribute,but it has not yet attracted the attention of researchers.Considering that the author of the conference paper is not necessarily present at the conference,for the convenience of the narrative,all the authors who published the paper at the same academic conference are considered to participate in the conference.Authors who co-occur in the same meeting tend to have the same research interests and are likely to belong to the same academic community,although there may not be a citation relationship between them.Therefore,the information of the authors participating in the conference will play a greater role in the accurate mining of the academic community.Information based on the co-attendance of the authors can form a network of author?s coattendance relationships.The co-authorship between authors is a closer link.The authors of co-authored thesis often come from the same or similar research fields,and are likely to belong to the same academic community,which will also play a role in the accurate mining of academic communities.Information based on the author’s co-authorship can form a network of author’s co-author relationships.The co-attendance network and the co-author relationship network are all a kind of relationship between the authors.It can be seen from the above that there are three relationships among authors: the citation relationship,the co-attendance relationship and the co-author relationship.These relationships can be used to form a multi-relationship diagram for the author.Weighting these three relationship strengths establishes an indicator of the degree of association between authors.Finally,using the author’s relevance to mine academic communities through community mining algorithms.Finally,using the author’s relevance to mine academic communities through community mining algorithms.The community mining algorithm used in this paper is an improved algorithm based on DBSCAN algorithm.The traditional DBSCAN algorithm needs to input the parameter by the user.Due to the limitation of personal experience,different operators may choose different parameters,in order to avoid experimental errors caused by personal factors.In this paper,an improved DBSCAN algorithm based on the statistical information of the data set itself is proposed as the clustering algorithm.This paper uses the DBLP database for experiments.The DBLP data record contains the title of the paper,author,year of publication,publication of journal or conference,reference and other information.Using this information,the author’s multi-relationship diagram consisting of the citation relationship,the co-attendance relationship and the co-author relationship is constructed.The author’s relevance is calculated by weighting the author’s three relationships.Finally,the association between these authors is used to conduct academic community mining through clustering algorithm.The experimental results show that the improved DBSCAN algorithm is better than the traditional DBSCAN algorithm for the academic community mining of the DBLP database.Co-attend factors and co-author factors help the mining of academic communities.The co-attend factors and co-author factors contribute to the stability of the cluster structure of the academic community..