微博数据获取与传播行为建模

发布时间：2018-03-11 07:28

本文选题：社交媒体　切入点：爬虫　出处：《华东师范大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着Web2.0应用的快速发展,社交媒体作为社会感知器用于记录人们日常生活、信息共享以及在线交友等,得到了商业界、政治界和学术界的广泛重视。分析社交媒体数据可以了解用户的关系网、理解用户的在线行为以及挖掘用户偏好,从而实现好友、产品以及服务的精准推荐；掌握信息在社交媒体中的传播途径和传播方式可以观测在线用户的集群行为,使虚拟空间成久安之势、建长治之业。因此,收集社交媒体数据,并感知用户在虚拟社区上的集群行为是重要而迫切的研究课题。由于社交媒体数据间具有强烈的依赖关系,传统的抽样技术不适用于收集社交媒体数据。另外,社交媒体数据的海量性、实时性和个性化等特点致使分析社交媒体数据成为一个非常有挑战的任务。而微博作为一类典型的社交媒体,它几乎具备当前社交媒体的所有特征。因此,本文以微博为载体研究了社交媒体数据的获取以及信息在社交媒体上的传播模型。主要贡献如下： ·设计和实现了一种基于微博社交网络结构的数据获取方法。根据网络中心化理论,选择可信的、正常的种子用户,通过微博平台的关注网络按照广度优先的策略设计了一个分布式爬虫,爬取并发布了海量微博数据；同时,根据爬取的数据集讨论了微博数据的实时获取。 ·给出了微博流行度的形式化定义,提出微博的流行度应该用转发量和可能被浏览次数这两种不同的方式进行度量。它们分别从实际受到影响进一步传播该信息的用户数和可能浏览到此信息的用户数两个角度度量了微博的流行度。 ·研究了微博的生命周期和引爆点。分析表明大多数具有高流行度的微博生命周期小于48小时；并且微博在传播过程中可能存在引爆点,即微博流行度会在信息传播过程中出现二次或二次以上的爆炸式增加。观察发现微博流行度随时间变化服从Sigmoid函数,文章利用Sigmoid函数拟合微博流行度随时间变化,同时给出估计模型参数的算法。实验验证了模型的有效性和参数估计的准确性。 ·建立了一个网上集群行为资源库,该资源库能以事件为单位,从时间、空间、情绪和传播网络这四个维度对事件进行可视化分析。该平台可以帮助人们了解事件的具体发展,事件参与者的大致组成、民众对事件的观点分布,事件的主要推动者等更进一步的信息。总的来说,本文以微博为载体,探索了基于社交网络结构获取社交媒体数据方法的可行性；基于收集的数据和流行度概念,对微博的传播方式进行了建模,研究微博生命周期和引爆点；最后建立社交媒体用户网上集群行为资源库,从多个维度展示社交媒体数据在研究用户集群行为中的作用。
[Abstract]:With the rapid development of Web2.0 applications, social media has been used as a social sensor to record people's daily lives, information sharing and online dating. The analysis of social media data can understand the user's network, understand the user's online behavior and excavate the user's preference, so as to realize the accurate recommendation of friends, products and services. How and how information is disseminated in social media allows you to observe the cluster behavior of online users, make the virtual space secure, and build a long-term business. Therefore, social media data are collected, And perceiving the cluster behavior of users in virtual community is an important and urgent research topic. Because of the strong dependency between social media data, traditional sampling techniques are not suitable for collecting social media data. The characteristics of real-time and personalization make it a challenging task to analyze social media data. Weibo, as a typical social media, has almost all the features of current social media. In this paper, Weibo is used as the carrier to study the acquisition of social media data and the dissemination model of information on social media. The main contributions are as follows:. 路designing and implementing a data acquisition method based on Weibo's social network structure. According to the theory of network centralization, we select trusted and normal seed users. Based on the focus network of Weibo platform, a distributed crawler is designed according to the strategy of breadth first, which crawls and publishes a mass of Weibo data. At the same time, according to the crawling data set, it discusses the real-time acquisition of Weibo data. 路give a formal definition of Weibo's popularity, It is suggested that the popularity of Weibo should be measured in two different ways, namely, the amount of forwarding and the number of times that the information may be browsed. They are further affected by the actual number of users who spread the information and the users who may browse to the information. Several two angles measure Weibo's popularity. 路studied Weibo's life cycle and tipping point. The analysis showed that the life cycle of the majority of those with high prevalence was less than 48 hours, and there might be a tipping point in the spread of Weibo. That is, Weibo's popularity will increase by two or more times in the process of information dissemination. It is found that Weibo's popularity varies with time from the Sigmoid function. The paper uses the Sigmoid function to fit the change of Weibo's popularity with time. At the same time, the algorithm of estimating the parameters of the model is given, and the validity of the model and the accuracy of the parameter estimation are verified by experiments. 路establish an online cluster behavior resource bank, which can visually analyze events in terms of events from the four dimensions of time, space, emotion and communication network. This platform can help people understand the specific development of events. The general composition of the event participants, the distribution of public views on the event, the main promoters of the event and other further information. In general, this paper takes Weibo as the carrier, explores the feasibility of acquiring social media data based on the social network structure, and builds a model for the transmission method of Weibo based on the collected data and the concept of popularity. The life cycle and tipping point of Weibo are studied. Finally, the social media users' online cluster behavior resource bank is established to demonstrate the role of social media data in the study of user cluster behavior from multiple dimensions.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092;G206

【共引文献】