Mining Of Relation Circle Based On Citation Network

The development of computer technology and information technology continues to accelerate,accompanied by the production of diverse big data technology,it is more and more widely used.In a wide range of data and large amount of data,it is difficult to quickly and accurately get the information wanted.Research paper is an important achievements of academic researchers,along with time changes the number of research papers increased dramatically,and citation between scientific research papers is becoming more and more frequent and complex,these citation form a very complex citation network.Different shapes and densities of the community structure are formed according to the reference relationship of the paper.In academic circles,it is usually divided into a small area,which will form a lot of smaller associations and will bring some obstacles to academic communication.Scholars outside the community want to learn the research results of other scholars in the corresponding field need to pay a higher expense.Therefore,an efficient method of community discovery can play an important role in the exchange of academic fields.Complex networks are composed of nodes which are transformed from complex systems,and the subset of nodes can form a community.The associations between nodes in the same community is much,and the associations between the groups are relatively small.Community finding in complex networks is closely related to clustering algorithms in data mining.In 2002,Girwan and Newman proposed GN algorithm,It is the most classical and scientific researchers well-known community discovery algorithm,belonging to a split hierarchical clustering algorithm.In 2004,an algorithm of community discovery based on information center degree proposed by Fortunato,It is the GN algorithm to remove the edge of the maximum number of edges modified to delete the information center of the largest edge.In 2005,Palla proposed the association discovery algorithm CPM(clique percolation method).It is the earliest community overlap algorithm,CPM algorithm is based on the complete subgraph,it appropriate the network which has more complete subgraph.Due to the close connection between the community internal node,the edge density is high and easy to form a clique.In the academic circles,there are different link strength between the citation.Most of the current community discovery algorithms ignore the link strength information between nodes.In the paper proposes an improved algorithm based on the research and analysis of the traditional density DBSCAN algorithm,and the improved algorithm is applied to the research of the academic circle.Experiments are performed on DBSCAN algorithm and improved algorithm using real data set DBLP.The results show that the improved algorithm proposed in this paper is better than the traditional DBSCAN clustering algorithm for the mining of academic circles.