基于模糊聚类算法的微博用户情感分析研究

发布时间：2018-05-18 11:05

本文选题：模糊聚类 + 网络爬虫　；参考：《西南石油大学》2014年硕士论文

【摘要】：随着Web2.0技术的发展,社交网站不断扩展其功能,人们可以在网上尽情地发表自己的观点,释放自己的情感。微博作为近几年来互联网的新兴产物,其情感计算与分析已成为计算机学、自然语言学、人类心理学等社会计算工作的重要研究课题。本文首先研究了模糊聚类相关理论知识,并将其应用于腾讯QQ空间说说的情感分析,为下一步将模糊聚类应用于微博用户情感分类分析提供了可操作性。模糊聚类分析首先要将文本数据数值化,对微博博文信息进行情感计算。因此本文在网络爬虫的技术原理上设计了新浪微博专用数据采集系统,建立了微博源数据库。利用基于知网(HowNet)基础情感词集的词语相似度计算方法计算微博情感词权值,建立了微博情感词库。在情感词库的基础上,分析了包含否定副词和程度副词等修饰词的微博短语情感强度值,并结合短句、短句与短句之间的关系,通过相应的情感值运算方法,计算得出整条微博消息的情感值。最后筛选了50名用户在10个不同时间段的情感值作为原始数据矩阵,利用模糊聚类算法将50名用户进行动态分类,并生成动态聚类图,由图可知根据不同的λ阈值,可以有不同的分类,并用F统计量检验法计算出最佳分类,再利用SPSS对分类结果预测作出每个分类用户的情感走势图,以便更直观地分析用户的情感变化。本文创新点主要在于：抓住人类情感是模糊不定的特征,利用模糊聚类来对已经通过情感计算的微博信息进行分类分析,政府、商家或企业可以根据不同的需求得到不同的分类结果,从而采取相应的措施。
[Abstract]:With the development of Web2.0 technology, social networking sites continue to expand their functions, people can express their views on the Internet and release their feelings. Weibo as a new product of the Internet in recent years, its emotional calculation and analysis has become an important research topic in computer science, natural linguistics, human psychology and other social computing work. This paper first studies the relevant theory of fuzzy clustering, and applies it to the emotional analysis of Tencent QZone, which provides the maneuverability for the next step to apply fuzzy clustering to the analysis of Weibo users' emotion classification. In fuzzy clustering analysis, the text data should be numerically valued, and the Weibo post information should be calculated with emotion. Therefore, this paper designs a special data acquisition system for Sina Weibo based on the technical principle of web crawler, and establishes the Weibo source database. The Weibo affective word weight is calculated by using the word similarity calculation method based on the basic affective word set of how Net. and the Weibo affective lexicon is established. On the basis of affective lexicon, this paper analyzes the affective strength of Weibo phrases, including negative adverbs and degree adverbs, and combines the relationship among short sentences, short sentences and short sentences. Calculate the emotional value of the entire Weibo message. Finally, the emotional values of 50 users in 10 different time periods are selected as the original data matrix, 50 users are dynamically classified by fuzzy clustering algorithm, and the dynamic clustering graph is generated. We can use F statistic test method to calculate the best classification, and then use SPSS to predict the classification result to make the emotion trend map of each classification user, so as to analyze the user's emotion change more intuitively. The innovation of this paper mainly lies in: grasping the feature that human emotion is fuzzy and indefinite, using fuzzy clustering to classify and analyze the Weibo information that has been calculated by emotion, the government, Businesses or enterprises can get different classification results according to different needs, and then take appropriate measures.
【学位授予单位】：西南石油大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.1;TP393.092

【相似文献】