基于有限信任的舆情数值建模与演化研究

发布时间：2018-05-26 10:55

本文选题：协程 + 网络舆情　；参考：《湖南科技大学》2014年硕士论文

【摘要】：近几年由于网络的发展，各种信息平台如天涯论坛，微博等的出现，加快了以往信息的传播速度。网络中的各种不同观点借由各种渠道开始迅速传播，这在方便沟通同时，亦易造成谣言、社会不良矛盾的扩散，甚至对于普通的事件进行添油加醋从而造成恶劣影响。有介于此，如何对于网络中的不良信息进行有效的控制，对正确的舆论进行有效的引导和扩散是一个急需解决的问题。由于网络中的信息是以文本表达，对于舆情分析的第一步需要对于文本信息进行数值建模，从而抽取出其中蕴含的舆情信息及相互之间的关系；其次需要对于用户进行建模，根据用户之间的交互来进行亲密度建模；最后通过演化模型对事件的发展进行预测和分析。有限信任模型考虑基本单位之间的交互关系，研究其亲密度、交互规则、演化规则、交互阈值等对于群体未来发展的影响。有限信任模型最初在统计物理方面显示出其优势，之后学者将其引入舆情演化的研究中，获得了比较好的效果，经过多年的研究，形成了几个比较典型的模型。Hegselmann-Krause模型（H-K模型）是其中的佼佼者，目前主要在仿真中取得了比较好的效果，但是在真实网络中，如何对于亲密度建模、交互舆情设定等，目前已有的研究还比较少。针对这些问题，本论文主要开展的工作如下： 1）利用基于协程的分布式爬虫框架爬取天涯数据，并对其进行数据建模及分析。首先介绍了协程的机制并实现了一个基于协程的网络爬虫框架，并详细介绍了在具体应用中的数据更新及信息去噪机制。通过对用户社区结构的分析，基于用户活跃度来对用户进行分类，并基于回复关系来构建活跃用户社区，最后利用PageRank来对用户进行影响力建模。通过查询扩展对论坛建立信息分布模型，通过对于事件抽取关键词，对其进行查询扩展，最终通过对于词频进行统计，构建信息舆论模型。 2）通过利用H-K模型的演化规则，基于粒子群的历史拟合方法对H-K模型的参数进行调优来对舆情演化进行预测。介绍了Sznajd模型与H-K模型的演化规则，并对粒子群算法进行了介绍，利用基于粒子群的网络拟合方法对历史舆情数据进行分析，通过基于粒子群方法的拟合来获取历史拟合参数，并利用演化数据进行修正从而获取演化模型，，实验证明采用历史拟合方法比利用固定值的方法能够获得更高的历史吻合率。最后通过实例分析来对我们方法进行介绍。 3）通过对于天卓舆情系统的设计分析，对数据库设计、架构设计进行了分析和并对相应舆情数据采集模块、舆情数值建模模块、舆情演化模块和舆情展示模块具体实现进行了分析。
[Abstract]:In recent years, due to the development of the network, the emergence of various information platforms such as Tianya Forum and Weibo has accelerated the speed of information dissemination in the past. All kinds of different viewpoints in the network begin to spread rapidly through various channels, which is easy to communicate, but also easy to cause rumors, the spread of bad social contradictions, and even to add oil to ordinary events, resulting in adverse effects. Therefore, how to effectively control the bad information in the network and guide and spread the correct public opinion is an urgent problem. Because the information in the network is expressed in the text, the first step of the analysis of public opinion needs to carry on the numerical modeling to the text information, so as to extract the public opinion information contained therein and the relationship between them. Secondly, it needs to model the user. Finally, the evolution model is used to predict and analyze the evolution of the event. The finite trust model considers the interaction between basic units, and studies the effects of affinity, interaction rules, evolution rules and interaction threshold on the future development of the population. The limited trust model initially showed its advantages in statistical physics, then the scholars introduced it into the research of the evolution of public opinion, and obtained a better result. After many years of research, The Hegselmann-Krause model (H-K model) is one of the best. At present, it has achieved good results in simulation, but in real networks, how to model affinity, set up interactive public opinion, etc. At present, the existing research is relatively small. In view of these problems, the main work of this paper is as follows: The main contents are as follows: 1) using the distributed crawler framework based on association, crawling the data of the horizon, and modeling and analyzing the data. Firstly, the mechanism of correlation is introduced and a network crawler framework based on correlation is implemented, and the mechanism of data updating and information de-noising in specific applications is introduced in detail. By analyzing the structure of the user community, the user is classified based on the user activity, and the active user community is constructed based on the response relationship. Finally, PageRank is used to model the influence of the user. The information distribution model of the forum is established by query expansion, the key words are extracted for the event, and the query extension is carried out. Finally, the information public opinion model is constructed through the statistics of word frequency. 2) by using the evolution rules of H-K model, the parameters of H-K model are optimized based on the historical fitting method of particle swarm optimization to predict the evolution of public opinion. The evolution rules of Sznajd model and H-K model are introduced, and the particle swarm optimization algorithm is introduced. The historical public opinion data are analyzed by using the network fitting method based on particle swarm optimization, and the historical fitting parameters are obtained by fitting based on particle swarm optimization method. The evolutionary model is obtained by modifying the evolution data. The experimental results show that the historical fitting method can obtain a higher historical coincidence rate than the fixed value method. Finally, through the analysis of examples to introduce our method. 3) through the analysis of the design of Tianzhuo public opinion system, the database design, the architecture design and the corresponding public opinion data collection module, the public opinion numerical modeling module, Public opinion evolution module and public opinion display module are analyzed.
【学位授予单位】：湖南科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.1;TP393.092

【参考文献】