Technology And Research Of The Academic Community Detecting Based On Paper Citation Relations

With the continuous development of Internet technology,the modern society becomes more and more digital, informational and networking. But the difficulty to obtain the information and the high dimension of information have already led to a big difficulty of various fields along with this big data era.In the academic field, due to geographical and other reasons, scholars of various research directions are usually isolated into small academic communities based on the meetings and university research teams,which in a large degree affected share,communication and inheritance of academic resources.Scholars outside the existing groups often have to pay huge cost in resource search and query if he or she wants to know some forefront academic resources of current areas and academic achievements of other professional scholars,which is much easier for some well-known experts and teams in the field.Therefore An effective academic community detecting method is of great importance in the process of promoting academic research and exchange.As an important carrier of work result record,communication and heritage of academic researchers in various fields,the academic paper reflects the author’s research level under the current research topic.The reference part reveals other academic groups and individuals that have same or similar research content and historical achievements.By studying the citation relationship of different papers,a huge paper citation network has set up. And then extracting the strength of association between different authors measured by quantity and also doing cluster analysis.It can effectively find academic communities.Paper citation relationship is divided into direct- indirect reference relationship,coupling relationship, the same citation relationship and joint work relationship.The direct-indirect references can be quantified by the reference path length in the network,while the coupling relationship, the same citation relationship need to be further processed to express the strength of association of the paper.And the quantitative index of the above several relations can complete the quantitative analysis of strength of the association of paper nodes from paper citation network.Thereby the quantitative analysis of the relationship between the nodes in the network is completed. By extracting the association relationship andthe relationship between the paper and the author, we can get a further quantitative representation of the author’s relationship network, and then by clustering we can get the academic community based on the citation relationship.The traditional clustering algorithms can be divided into clustering algorithm based on hierarchical, partition, density and network and so on. For the complex network of the authors,such an complex communities detecting work which has complex structure and inconsistent classification shape. the clustering algorithm based on density can get a better result. But due to the fix in the choice of neighborhood radius Eps,the traditional DBSCAN clustering algorithm cannot achieve better results in clustering of extreme points like the edge and the core.In this paper,by studying the traditional DBSCAN algorithm, a new noise iterative clustering algorithm based on Eps parameter set is proposed to improve the traditional DBSCAN algorithm. The improvement effect of clustering is verified by the external evaluation index of real data setSonar. At the same time, the improved algorithm is applied to the discovery process of academic communities.Also,respectively use S(c) and Ocq as internal and related evaluation index to verify and detect the academic communities and further demonstrate the effectiveness of this algorithm.