期刊名称：International Journal of Hybrid Information Technology
印刷版ISSN：1738-9968
出版年度：2015
卷号：8
期号：7
页码：399-410
DOI：10.14257/ijhit.2015.8.7.36
出版社：SERSC
摘要：This paper introduces a new multi-view and controllable community-uncovering algorithm, an achievement of improving PageRank algorithm and Spin-glass model, which can avoid the overlapping community structure in the process of detecting communities by means of other algorithms and also helps to improve the usual community-expansion model. The process of uncovering communities by the introduced algorithms can be divided into three steps: first, identifying the nuclear one among nodes ranked by the advanced PageRank algorithms; Second, through using multi-view recognition modularity provided by Potts spin-glass model, optimizing the expansion model of local community that is found by applying the improved Iterative Greedy algorithm to eliminate the traditional modularity's limits in the resolution limit and the following negative effects. Finally, grasping the overlapping structure and notes carefully. By analyzing and comparing the two results of respectively using PRSGMFCA and traditional technical schemes in both computer simulation network and the real network, it proves that the former enjoys stronger stability and higher accuracy than the latter, and its computation complexity is also acceptable.
