基于相似性模块度的层次聚合社区发现算法

发布时间：2018-07-05 06:26

  本文选题：Fast + Unfolding算法　； 参考：《系统仿真学报》2017年05期

【摘要】：Fast Unfolding是一种基于模块度优化的层次聚合社区发现算法,其优点是运行速度很快,不足之处是准确度有待提升,这是因为该算法采用传统模块度作为合并社区的衡量标准,而传统模块度函数在计算时只考虑节点间的链接信息,忽略邻居节点的影响,导致会出现两个节点共同邻居较多但由于节点间链接信息较弱不能被合并的情况,从而影响结果的准确度。针对该不足之处,通过引入优化后的相似度来改进Fast Unfolding算法的模块度函数,提出一种基于相似性模块度的层次聚合社区发现算法,并采用归一化互信息量即NMI(Normalized Mutual Information)作为评价算法准确性的指标,在真实网络和LFR(Lancichinetti Fortunato Radicchi)人工合成网络上进行实验,结果表明改进算法检测社区结构的准确度有明显改善。
[Abstract]:Fast merging is a hierarchical aggregate community discovery algorithm based on modularity optimization, which has the advantages of fast running speed and high accuracy. This is because the traditional modular degree is used as the measure of merging community. However, the traditional modular degree function only considers the link information between nodes and neglects the influence of neighbor nodes, which leads to the situation that two nodes have more common neighbors but the link information between nodes can not be merged because of the weak link information between nodes. Thus, the accuracy of the results is affected. In order to solve this problem, a hierarchical aggregation community discovery algorithm based on similarity modularity is proposed by introducing the optimized similarity to improve the modularity function of Fast portfolio algorithm. NMI (Normalized Mutual Information) is used as an index to evaluate the accuracy of the algorithm. The experiments are carried out on real network and LFR (Lancichinetti Fortunato Radicchi) artificial synthetic network. The results show that the accuracy of the improved algorithm in detecting community structure is obviously improved.
【作者单位】： 中国矿业大学计算机学院;
【基金】：国家自然科学基金(61402482) 中国博士后基金(2015T80555) 江苏省博士后基金(1501012A)
【分类号】：O157.5;TP301.6
，

本文编号：2099286