基于概率因子模型的演化社会网络分析方法研究

发布时间：2018-05-13 12:15

本文选题：演化社会网络 + 概率因子模型　；参考：《厦门大学》2014年硕士论文

【摘要】：演化社会网络是动态更新的社会网络。随着信息技术的发展,信息交换的成本迅速下降,大量易用的通信、互联网交互平台迅速发展。而在这类平台上所构建的社会网络通常具有高度动态的结构,并且其内部的结构特征相比传统网络更加丰富。演化社会网络的分析目前依然处于起步阶段,但已有的大量静态网络以及图论的研究可以被演化社会网络分析借鉴或进一步扩展。本文在对已有的分析模型研究基础上,对演化社会网络分析方法进行了理论探索和实践。本文主要针对演化社会网络分析的框架,网络变化检测以及社团划分问题进行了研究,主要工作集中在以下方面： 1、提出了新的演化社会网络分析框架,利用因子模型优化关联强度特征向量,重建关联强度矩阵,并通过计算演化指标捕获社会网络的结构变化点的方法。 2、定义了概率视角的演化社会网络,在此基础上提出了演化社会网络的概率因子模型,并将模型用于网络结构变化点的检测。 3、提出了基于概率的演化社会网络社团划分的方法。沿用相关聚类的启发式算法,在演化社会网络的分析框架之下对社团进行划分和评价。上述理论研究在随机生成的数据集以及真实数据集进行了实践。实验结果表明Event-Detect算法以及Prob-Event-Detect算法有效的检测出了演化网络的结构变化。通过对比已有的社团划分算法,Prob-Cluster-Detect算法亦表现出较好的鲁棒性以及社团划分的合理性。本文最后部分对工作进行了总结。此外在研究展望部分还分别针对演化社会网络的时空模型、演化社会网络的关联强度模型以及基于关系的演化社会网络分析方法进行了讨论,给出在本文所给出的框架之下切实可行的若干工作方向。
[Abstract]:Evolutionary social network is a dynamic updating social network. With the development of information technology, the cost of information exchange decreases rapidly, a large number of easy-to-use communications, the rapid development of Internet interaction platform. The social network constructed on this platform usually has a highly dynamic structure, and its internal structural characteristics are more abundant than the traditional network. The analysis of evolutionary social networks is still in its infancy, but a large number of static networks and graph theory can be used for reference or further expanded by evolutionary social networks. In this paper, based on the research of the existing analytical models, the analytical methods of evolutionary social networks are explored in theory and put into practice. This paper focuses on the framework of evolutionary social network analysis, network change detection and community division. The main work is focused on the following aspects: 1. A new analytical framework of evolutional social network is proposed. The eigenvector of association strength is optimized by factor model, the matrix of association strength is reconstructed, and the structural change point of social network is captured by calculating the evolution index. 2. The probabilistic factor model of evolutionary social network is proposed based on the definition of evolutionary social network from the point of view of probability, and the model is used to detect the change point of network structure. 3. A probabilistic method of community division in evolutionary social networks is proposed. Using the heuristic algorithm of correlation clustering, the community is divided and evaluated under the analytical framework of evolutionary social network. The theoretical research is carried out in random data sets and real data sets. The experimental results show that the Event-Detect algorithm and the Prob-Event-Detect algorithm can effectively detect the structural changes of the evolutionary network. Compared with the existing community partition algorithm, Prob-Cluster-Detect algorithm also shows good robustness and the rationality of community partition. The last part of this paper summarizes the work. In addition, the spatio-temporal model of evolutionary social network, the correlation strength model of evolutionary social network and the analytical method of evolutionary social network based on relationship are discussed respectively in the part of research prospect. Some feasible working directions under the framework given in this paper are given.
【学位授予单位】：厦门大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.02

【参考文献】