【作者】 李剑；

【导师】 闫光辉；

【作者基本信息】 兰州交通大学 ， 计算机应用技术， 2015， 硕士

【摘要】 微博作为一种新兴的社交平台,在社交网络中占据了十分重要的地位。由于微博有信息传播迅速、实时性强、交流方便等特点,因此它吸引了大量来自不同行业、领域的用户。正是由于微博用户职业的多种多样,兴趣的多种多样等,导致了在微博网络中存在许多不同种类的社区。微博社区发现的主要目的是通过采用某种方法,对微博用户网络进行组织与划分,以使相互之间有最大相似度的微博用户聚集在一起,从而形成一个个的微博社区。显然,可将数据挖掘技术应用于微博网络,以达到发现微博网络中的社区结构的目的。该研究领域业已存在相当数量的社区模型和社区发现算法,并取得了一定的研究成果。随着互联网技术的发展和社交工具的日益推广应用,人们的社交方式也在逐渐改变,对社区发现研究提出了新的挑战。基于以上背景,本文在分析已有微博社区发现技术的基础上,针对传统算法大多仅考虑单一因素,难于有效识别微博中多个社区的问题,借鉴社会网络分析思想,综合考虑微博用户的主题相似度和链接关系,基于标签传播算法展开微博社区发现的研究,主要研究内容为:(1)以新浪微博为例,分析研究了微博的类型、结构特征与功能,分析了微博社区的无标度性、链接关系与同质性。(2)研究了社区发现的经典技术与文本分析的基本方法,并进行了对比分析。(3)针对微博这一被广泛使用的社交媒体,在将其抽象成为一个网络之后,以标签传播算法为基础,通过分析网络中节点的主题相似度及不同节点之间的标签传播概率,提出了一种从复杂网络分析角度出发进行研究的微博社区发现算法,并针对新浪微博数据集,与GN算法和Spectral Bisection Method进行了初步的对比实验,实验证明了本方法的有效性。更多还原

【Abstract】 Microblog, as a new kind of social network platform, plays a significant role in China social networks. The microblog attracts amount of people due to the information’s rapidly diffuse, the real-time communication and the convenient of interaction. As the microblog users have different occupations or interests, the microblog network exists many different kinds of community.The purpose of microblog community discovery is to find the communities in a microblog network in which the users have the largest similarity among them. Obviously, techniques in data mining can be used in the network to detect the community structures in it. Scholars at home and abroad do lots of research about it and achieve some results. But with the development of internet and social network tools, people’s way of communication is changing, which makes many conventional algorithms be not suitable for new social network tools any longer. Conventional community discovery algorithms are generally based on either links or interests and don’t take limits of obtaining microblog users’ social information into consideration. As a result, they can’t detect multiple communities effectively. In such case, based on the analysis of existing algorithms of web communities discovery, this thesis focuses on microblog community discovery based on Label Propagation Algorithm. The main contents of this thesis are as follow:(1) Use Sina Weibo as an example, we discuss the microblog’s types, structure features and functions. Moreover, we also introduce the existing web community model, scale-free property, links and homogeneity.(2) We introduce and analyze the classical technologies of community detection and some basis methods for text mining.(3) For microblog, which is a widely used social media, we use the Label Propagation Algorithm as the foundation, after have an analysis of the topic relevancy between the nodes in a network and the probability of label propagation between the nodes, we propose an algorithm for microblog community discovery from the perspective of complex network analysis. Then do some experiments on a real Sina Weibo dataset. We compare our algorithm with two existed algorithms(GN Algorithm and Spectral Bisection Method). The experimental results show that our algorithm has a good performance on community discovery.更多还原