线上社交网络信息传播的研究与实现

发布时间：2018-04-25 12:26

本文选题：转发概率 + 在线社交网络　；参考：《西安电子科技大学》2014年硕士论文

【摘要】：随着社交网络平台的迅速发展,人们在社交网络上的活动越来越多,以国内新浪微博为例,在2013年微博用户数已达到6亿。因此越来越多的研究者将在线社交网络作为研究对象。在在线社交网络中,用户可以讨论自己的想法,发表自己的意见,表达自己的兴趣等,所有这些行为产生了大量的社交数据。其中,如何对信息在整个网络上的传播进行模拟成为一个热点研究话题。转发行为是组成信息传播的原子行为。因此,本文先从影响转发行为的因素出发得出转发概率,然后提出了基于转发行为的信息传播模型,最后在真实的数据集中对信息传播进行了模拟。实验表明,提出的模型能够克服传统模型中固定转发概率带来的同质性,能够更好地模拟真实信息传播过程。具体工作如下:微博数据的抓取。本文搭建了基于Hadoop的分布式微博数据抓取平台。在八台机器上,对微博的用户数据、关系数据和文本内容数据进行了抓取。转发行为的分析。通过对微博数据分析来选择合适的特征和相应的模型。利用相应的特征和模型,得到用户之间的转发概率,本文对每个用户采用了逻辑回归模型来进行数据拟合,首先对数据进行预处理,提取出相应的特征,将处理后的数据输入逻辑回归模型进行训练,最后根据模型得出每个用户的转发概率。信息传播模型的模拟。首先,在微博数据中抽取出网络结构。由于用户之间的转发组成了信息的传播,根据转发行为的分析,本文提出了pSIS?与pIC?传播模型。最后,在提取的网络中进行了传播模拟的实验。通过实验,本文发现了一些新的现象:在传播模型中,拥有固定转发概率的信息传播虽然在传播范围上跟真实社交网络中相同,但其传播速度明显低于真实网络中的信息传播。另外,初始信息发布人的选择对于信息传播也是非常重要的,经过研究表明某种程度上这种选择可以使我们做出有效的策略来控制谣言的传播与舆情监控。
[Abstract]:With the rapid development of social networking platforms, people are becoming more and more active on social networks. Taking Sina Weibo as an example, the number of subscribers reached 600 million in 2013. Therefore, more and more researchers take the online social network as the research object. In online social networks, users can discuss their ideas, express their opinions, express their interests and so on, all of which produce a lot of social data. Among them, how to simulate the spread of information on the whole network has become a hot topic. Forwarding behavior is the atomic behavior that constitutes the transmission of information. Therefore, this paper first obtains the forwarding probability from the factors that affect the forwarding behavior, then puts forward the information transmission model based on the forwarding behavior, and finally simulates the information transmission in the real data set. Experiments show that the proposed model can overcome the homogeneity brought by the fixed forwarding probability in the traditional model and can better simulate the process of real information transmission. The specific work is as follows: Weibo data capture. This paper builds a distributed Weibo data grab platform based on Hadoop. In eight machines, Weibo's user data, relational data and text content data were grabbed. Analysis of forwarding behavior Through the Weibo data analysis to select the appropriate characteristics and the corresponding model. Using the corresponding features and models, the forwarding probability between users is obtained. In this paper, the logical regression model is used to fit the data for each user. Firstly, the data is preprocessed to extract the corresponding features. The processed data input logical regression model is trained and the forwarding probability of each user is obtained according to the model. Simulation of information transmission model. First, the network structure is extracted from Weibo data. Since forwarding between users constitutes the spread of information, according to the analysis of forwarding behavior, this paper proposes a pSIS? With PICs? Propagation model. Finally, the propagation simulation experiment is carried out in the extracted network. Through experiments, we find some new phenomena: in the transmission model, the information transmission with fixed forwarding probability is the same as that in the real social network, but its propagation speed is obviously lower than that in the real network. In addition, the choice of initial information publisher is also very important for information dissemination. The research shows that to some extent, this choice can enable us to make effective strategies to control the spread of rumors and public opinion monitoring.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【相似文献】