社交网络中基于信任模型的社区发现算法研究

发布时间：2018-06-13 06:36

本文选题：社交网络 + 重叠社区发现　；参考：《合肥工业大学》2017年硕士论文

【摘要】：随着互联网技术的快速发展,社交网络已逐渐成为了人们日常交友沟通、个人生活展示及消息发布的主要平台。社区发现是社交网络研究中的一个热点,挖掘社交网络中潜在的社区结构有助于深入理解网络的拓扑结构特点,也能为舆情监测、意见领袖发现和个性化推荐等诸多方面的研究与应用提供有力的支持。但目前随着社交网络规模的不断增大,如何从愈发复杂的社交网络中简单高效地挖掘出具有潜在特征的重叠社区结构成为了一项具有挑战性的问题。同时,社交网络中的用户之间不仅存在着显性关系,还存在好友相似、属性相似和兴趣相似等形式所表现出的隐性关系。为了更加合理地分析社交网络中用户之间的关系特征,可以通过使用信任来衡量用户间的个体权重、关系强度等关系属性,并通过定义信任的计算方法与传递规则来完成社交网络中的关系描述,从而能够有效提升社交网络分析的准确性因此,为解决已有社区发现算法中存在的问题,本文首先定义了一种社交网络中节点间信任的计算方法,通过使用信任来描述节点之间的关系特征,并在此基础之上提出了基于节点间信任的重叠社区发现算法,最后通过对比实验完成了算法的验证。具体的研究内容如下:1)提出了一种社交网络中融合了节点间关系强度与相似性的信任计算方法。在相关信任计算研究的基础之上,分析与选择社交网络中影响节点间信任的因素之后,针对由节点间关系强度产生的关系信任和节点间相似性产生的相似信任,分别给出了对应的计算方法。社交网络环境中信任的计算方法是本文后续研究的重要基础。2)设计了一种社交网络中基于节点间信任的局部重叠社区发现算法TLCDA(Trust-Based Local Overlapping Community Detection Algorithm)。TLCDA算法将社交网络抽象成为数据场后使用信任势来描述局部范围内节点之间的影响作用,并通过使用粗糙K-Mediods聚类完成重叠社区发现。3)制定了实验方案并完成对比分析。本文选取了LFR人工基准网络、经典真实网络和微博网络三种不同类型的网络,给出了社区发现的效果评价指标,并通过与经典的社区发现算法进行对比完成了TLCDA算法的效果验证。
[Abstract]:With the rapid development of Internet technology, social network has gradually become the main platform for people to make friends and communicate, personal life display and news release. Community discovery is a hot topic in the research of social network. Mining the potential community structure in social network can help to understand the topological characteristics of the network and monitor the public opinion. Research and application of opinion leader discovery and personalized recommendation provide strong support. However, with the increasing scale of social network, how to find the overlapping community structure with potential characteristics from the increasingly complex social network has become a challenging problem. At the same time, there are not only dominant relationships among users in social networks, but also hidden relationships in the form of similar friends, similar attributes and similar interests. In order to analyze the relationship characteristics of users in social network more reasonably, we can use trust to measure the individual weight and relationship strength among users. The relationship description in social network can be completed by defining trust calculation method and transfer rule, which can effectively improve the accuracy of social network analysis. Therefore, in order to solve the problems existing in existing community discovery algorithms, This paper first defines a computing method of trust between nodes in social networks, describes the relationship characteristics between nodes by using trust, and then proposes an overlapping community discovery algorithm based on trust between nodes. Finally, the algorithm is verified by contrast experiment. The main contents of this paper are as follows: (1) A trust computing method which combines the strength and similarity of the relationship between nodes in social networks is proposed. Based on the research of related trust computing, this paper analyzes and selects the factors that affect the trust between nodes in social network, and then analyzes the relationship trust generated by the strength of the relationship between nodes and the similarity between nodes. The corresponding calculation methods are given respectively. Trust computing method in social network environment is an important foundation of the following research in this paper. (2) A local overlapping community discovery algorithm based on trust between nodes in social network is designed. TLCDA Trust-Based Local overlapping Community Detection algorithm. TLCDA algorithm abstracts social network. After becoming a data field, a trust potential is used to describe the effects between nodes in a local scope. By using rough K-Mediods clustering to complete the overlapping community discovery. 3) the experimental scheme was developed and the comparative analysis was completed. In this paper, we select three different types of networks: LFR-artificial benchmark network, classical real network and Weibo network, and give the evaluation index of community discovery effect. The effect of TLCDA algorithm is verified by comparing with the classical community discovery algorithm.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP301.6;TP393.09

【相似文献】