基于社交网络的舆情关键技术研究

发布时间：2018-01-12 13:36

本文关键词：基于社交网络的舆情关键技术研究　出处：《吉林大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着移动互联网的深入发展以及5G时代的到来,各类社交应用持续发展,互联网平台呈现泛社交化。社交网络中的典型代表微信、微博,在舆情监测分析中均扮演着重要的角色。相较于微博平台舆情监测技术的日渐成熟,基于微信平台的相关研究较为匮乏,因此,本文重点研究了基于微信平台的舆情监测分析关键技术。本文根据社交网络的舆情系统需求,设计系统总体架构,解决微信平台公众号数据获取问题,并对网络舆情进行倾向性分析。本文提出word2vec模型与机器学习相结合的微信文本倾向性分析模型,能够有效提高分类准确率、精确度、召回率。本文的主要工作如下:(1)设计了微信平台舆情监测分析系统框架,并对框架内的各功能模块逐一设计实现。阐述了系统任务执行流程,设计并实现数据采集功能、数据存储功能、文本倾向性分析功能、舆情报告展示功能。(2)通过Python爬虫及Hook技术相结合的方式,实现微信平台公众号数据的自动获取,并存储于服务器MySQL数据库中。在数据采集模块中详细说明了各组件工作原理及采集流程,明确移动终端组件、服务器端组件构成及通信方式,利用多个移动终端采集,提高采集速率。信息通过去噪处理后存储,为舆情分析提供数据支持。(3)本文在研究特征选择算法及分类算法的基础上,进一步研究了word2vec模型,提出机器学习方法与word2vec相结合的微信文本倾向性分析模型。并将提出的模型与传统分类模型作对比研究,通过模型调优,使其在微信文本倾向性分类中效果更好。实验结果表明,本文提出的模型,在微信平台文本倾向性分析中表现良好。准确率达到84%,精确度达到86%,召回率达到83%,对比传统分类模型分别提升5%、7%和5%,ROC曲线AUC值达到0.88,总体分类性能得到提升。本文的主要贡献和创新点如下:(1)提出了机器学习与word2vec相结合的微信数据舆情分析模型,利用机器学习调优方法,对模型进行实验训练,使得系统分类性能整体得到提高,并实现微信文本的倾向性分析。(2)设计并实现了微信平台的数据采集功能,利用Python爬虫和Hook技术相结合的方式,解决了传统网页爬虫访问服务器频率受限的问题。通过多个移动终端并行采集,提高了数据采集速率,实现了微信数据的大规模获取,为后续实验研究提供数据支持。
[Abstract]:With the further development of mobile Internet and 5G era, all kinds of social sustainable development, the Internet platform has Pan social. Social networks are the typical representatives of WeChat, micro-blog, in the public opinion monitoring and analysis play an important role. Compared to the micro-blog platform and monitoring technology, based on the related research of WeChat platform is scarce, therefore, this paper focuses on the research of public opinion monitoring WeChat platform. Based on the analysis of key technology based on the needs of the public opinion system of the social network, the overall system architecture design, solve the problem of obtaining WeChat platform public number data, and public opinion on the network orientation analysis. In this paper, WeChat text orientation word2vec model and machine learning the combination analysis model, can effectively improve the classification accuracy, precision and recall. The main work of this paper are as follows: (1) the design of WeChat Taiwan public opinion monitoring and analysis system, and the functional modules of the design and implementation of the framework. One describes the system task execution process, the design and implementation of data acquisition, data storage, text orientation analysis, public opinion reports show. (2) through the combination of Python and Hook crawler technology combination, realize the automatic get the WeChat platform public number data, and stored in the MySQL server database. The data acquisition module in detail the various components and working principle of the acquisition process, clear mobile terminal components, server components and communication mode, using multiple mobile terminal acquisition, increasing the acquisition rate. Through denoising after storage. Provide data support for the analysis of public opinion. (3) the basic algorithm and classification algorithm in the study of characteristics, further study of the word2vec model, a machine learning method WeChat text orientation combined with word2vec analysis model. Compared with the traditional model and proposes the classification model, through the model optimization, the effect is better in the WeChat text orientation classification. The experimental results show that the proposed model performs well in the WeChat platform text orientation analysis. The accuracy rate of reached 84%, precision 86%, recall rate reached 83%, 5% improve the traditional classification and comparison model respectively, 7% and 5%, the ROC curve of AUC value reached 0.88, the overall classification performance has been improved. The main contributions and innovations of this paper are as follows: (1) proposed a model of WeChat data public opinion machine learning combined with word2vec analysis. Using machine learning optimization method, experimental training on the system model, the overall classification performance is improved, and the tendency analysis of WeChat text. (2) the design and implementation of data acquisition of WeChat platform The function, using the combination of Python and Hook crawler technology combination, to solve the traditional web crawler server access frequency is limited. Collected in parallel by the plurality of mobile terminals, improves the data acquisition rate, to achieve a large-scale acquisition of WeChat data, provide data support for subsequent experimental research.

【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09;TP181

【相似文献】