基于微博社会网络的用户兴趣模型研究

发布时间：2018-05-20 04:24

本文选题：微博 + 复杂网络　；参考：《上海交通大学》2013年硕士论文

【摘要】：微博是一个基于用户关系的信息分享、传播以及获取平台，是基于web2.0平台新近兴起的开放互联网社交服务。在微博平台中，用户通过关注形成人际关系网，用户发布的信息通过转发等方式以病毒的方式飞速传播。这使得微博不仅仅是扩大人际圈实现社会交往的社交网络，更是获取最新资讯和各方评论信息的重要媒介。近年来国外基于社会网络的用户兴趣研究越来越多，利用用户兴趣模型提供的精准化广告投放服务成本低、效果好，逐渐成长为一种主要盈利模式。国内目前还没有十分完善成熟的类似系统，因此本文针对微博的社会网络特性进行了如下研究：第一，分析了微博社会网络的信息结构，概括了微博用户兴趣的构成，针对传统微博信息采集方案的不足提出了具有可行性的微博信息采集系统设计方案，实现的分布式采集系统在实际应用中效率和扩展性好于传统采集方案；第二，构建出真实的微博社会网络关系，结合复杂网络中节点重要度算法相关理论，以PageRank算法思想为基础提出了微博用户重要度的收敛算法，以用户重要度反映微博用户的现实影响力；第三，研究了微博短文本的关键词提取方案，包括微博中冗余信息的过滤，利用自然语言处理工具分词，用同义词词林计算词语相似度，根据词语间相似度关系构造无向图，用图的邻接矩阵计算词语重要度，从而提取出微博文本中的关键词；第四，在用户重要度和微博关键词的基础上，，提出了基于TF-IDF改进的微博用户兴趣特征项提取算法，利用向量空间模型建立微博用户的兴趣空间，通过长期兴趣与短期兴趣定义用户的兴趣度，完成用户兴趣建模，并通过实验证明了该模型与人工标注的结果相吻合且具有良好的准确性。
[Abstract]:Weibo is an information sharing, dissemination and access platform based on user relationship. It is an open Internet social service based on web2.0 platform. In the Weibo platform, the user forms the interpersonal network through the concern, and the information released by the user spreads rapidly through the way of transmission and so on. This makes Weibo not only a social network to expand the interpersonal circle to realize social interaction, but also an important medium to obtain the latest information and comments. In recent years, there are more and more researches on user interest based on social network abroad. The precision advertising service provided by user interest model has low cost and good effect, and has gradually become a major profit model. At present, there is no mature similar system in China, so this paper studies the characteristics of social network of Weibo as follows: Firstly, this paper analyzes the information structure of Weibo social network, generalizes the composition of Weibo user's interest, and puts forward a feasible design scheme of Weibo information collection system aiming at the shortage of traditional Weibo information collection scheme. The efficiency and expansibility of the distributed acquisition system is better than that of the traditional collection scheme. Secondly, the real Weibo social network relationship is constructed. Combined with the related theory of node importance algorithm in complex network, the convergence algorithm of Weibo user importance is proposed based on PageRank algorithm. The influence of Weibo users is reflected by the importance of users. Thirdly, the keyword extraction scheme of Weibo short text is studied, including the filtering of redundant information in Weibo, the use of natural language processing tools to segment words, the calculation of word similarity by using synonym forest, and the construction of undirected graph according to the similarity relationship between words. The importance of the words is calculated by the adjacency matrix of the graph, and the keywords in the Weibo text can be extracted. Fourthly, on the basis of user importance and Weibo keyword, an improved Weibo user interest feature extraction algorithm based on TF-IDF is proposed, and the interest space of Weibo user is established by using vector space model. The user's interest degree is defined by long-term interest and short-term interest, and user interest modeling is completed. Experiments show that the model is consistent with the results of manual annotation and has good accuracy.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP393.092

【参考文献】