面向DBWorld数据挖掘的学术社区发现算法

Academic community detection algorithm for DBWorld data mining

针对传统社区发现算法多数是基于单一关系的同构学术社会网络,而包含多种关系的异构学术网络社区发现算法还不多的情况,提出一种基于FCM(fuzzy C-means)和结构洞的学术社区发现算法——HAFCD算法。从构建基于DBWorld邮件数据的异构学术网络出发,通过分析异构网络中的多种关联关系和节点内容的相似性,提出改进的语义路径模型,计算评审人间的相似度。基于此,该算法根据结构洞越少、网络闭合性越高这一事实,将结构洞理论融入FCM算法进行异构学术社区发现。通过与现有的谱聚类和路径选择聚类算法进行实验比较表明,该算法具有较好的计算效果。

In view of the most of traditional community discovery algorithms were based on the single relationship between the isomorphism of academic social network, and there were few heterogeneous network academic community discovery algorithm which contained a variety of relations, this paper put forward a kind of academic community discovery algorithm based on FCM and structural hole, called HAFCD algorithm. This paper was based on heterogeneous academic network which from the DBWorld e-mail data, then through the analysis of the multiple correlation and nodes content’s similarity in heterogeneous network, put forward an improved semantic path model to compute the similarity between the assessor. Based on this, according to the fact that the less structural holes, the higher the network closed, the HAFCD algorithm found heterogeneous academic community by applying the theory of structural holes into FCM algorithm. Through the comparison of HAFCD algorithm with the spectral clustering algorithm and PSC algorithm, the proposed algorithm has better effect on community detection.