移动社交网络相依关系及社区发现算法研究

发布时间：2018-02-23 05:30

本文关键词： 移动社交网络相依网络鲁棒性社区发现　出处：《哈尔滨工业大学》2014年硕士论文　论文类型：学位论文

【摘要】：本文探讨了移动社交网络的结构特征和网络特性,基于复杂网络理论提出了一种基于组增长的无标度网络模型,根据移动社交网络中用户和设备之间不同的依赖支持关系构建了两种不同网间关系的相依网络,一种是描述移动社交网络一对一相互依赖关系的相依网络,另一种是描述移动社交网络多重依赖支持关系的相依网络。文中将攻击策略分为随机攻击和蓄意攻击度值大的节点,分别讨论了在两种攻击策略下不同相依关系的移动社交网络的结构变化,并且分析了它们的鲁棒性。通过理论推导得到在随机攻击下,不同网络模型发生大规模失效现象的阈值,然后进行计算机仿真模拟相继故障发生的过程,在去除一定比例节点之后,网络节点发生相继失效,最终剩余的未失效的节点组成了剩余最大聚簇,剩余最大聚簇的大小代表了网络鲁棒性的强弱。当去除节点比例达到阈值时,网络发生“雪崩”现象,即不存在剩余最大聚簇。研究结果表明,在随机攻击下,具有相依关系的移动社交网络模型比单个网络模型的鲁棒性弱;在蓄意攻击下,具有相依关系的移动社交网络模型比单个网络模型的鲁棒性强。然而,无论是在随机攻击还是蓄意攻击下,具有一对一相依关系的移动社交网络和具有多重依赖支持关系的移动社交网络相比鲁棒性都要弱。根据真实社交网络的动态性,本文提出了一种自适应社区发现算法,与传统的静态社区发现算法不同的是该算法引入了自适应的概念,不需要考虑当前网络全部拓扑结构,只需通过之前的网络社区划分和网络结构的变化就能划分出新的社区结构。该算法可以在动态网络中进行社区划分,本文通过在真实数据集和人工合成数据集上的实验分析该算法的准确性。实验结果表明,该算法在NMI评价标准下,与其他算法相比具有较好的表现。文章的最后我们对研究工作做出了总结,分析研究中存在的不足之处,提出未来的研究展望。
[Abstract]:In this paper, the structure and network characteristics of mobile social networks are discussed, and a scale-free network model based on group growth is proposed based on complex network theory. According to the different dependency support relationships between users and devices in mobile social networks, two kinds of dependent networks with different network relationships are constructed, one is the dependent networks that describe the one-to-one interdependence of mobile social networks. The other is a dependent network that describes the multi-dependency support relationship of mobile social networks. In this paper, the attack strategy is divided into nodes with high degree of random attack and deliberate attack. The structural changes of mobile social networks with different dependencies under two attack strategies are discussed, and their robustness is analyzed. The threshold of large-scale failure occurs in different network models, and then computer simulation is carried out to simulate the process of successive failures. After removing a certain proportion of nodes, the network nodes fail successively. The residual nodes formed the largest residual cluster, and the residual maximum cluster size represented the robustness of the network. When the proportion of nodes removed reached the threshold, the "avalanche" phenomenon occurred in the network. The results show that under random attack, the robustness of the dependent mobile social network model is weaker than that of the single network model, and the robustness of the mobile social network model is weaker than that of the single network model under the deliberate attack. The model of mobile social network with dependency is more robust than the model of single network. However, whether in random attack or deliberate attack, The robustness of mobile social networks with one-to-one dependencies and multi-dependency support relationships is weaker than that of mobile social networks. According to the dynamic nature of real social networks, an adaptive community discovery algorithm is proposed in this paper. Different from the traditional static community discovery algorithm, this algorithm introduces the concept of adaptive, and does not need to consider all the current network topology. The new community structure can be divided by the former network community division and the network structure change. This algorithm can divide the community in the dynamic network. In this paper, the accuracy of the algorithm is analyzed by experiments on real data sets and synthetic data sets. The experimental results show that the algorithm is based on the NMI evaluation standard. At the end of the paper, we summarize the research work, analyze the shortcomings of the research, and put forward the future research prospects.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.01;O157.5

【相似文献】