校园BBS舆情监测分析研究

发布时间：2018-04-11 10:48

本文选题：校园 + BBS　；参考：《福州大学》2014年硕士论文

【摘要】：随着互联网的发展,校园BBS已经成为了大学生网上交流的主要平台之一通过校园BBS舆情监测,可以从校园BBS中采集数据并对数据进行分析处理,根据分析结果进行准确的校园舆情状态控制、校园舆情发现与预警等相关操作。本文分析了校园BBS舆情监测的整体组织框架,将校园BBS舆情分析系统分成四个模块,分别是论坛数据采集模块、数据预处理模块、舆情信息分析模块和舆情监测与服务模块。本文分别介绍了在这四个模块中用到的一些常用的技术,主要有基于校园BBS的聚焦爬虫算法和改进技术、针对BBS逻辑结构分析的帖子去噪技术,以及中文分词技术,停用词技术、文本分类方法、关键词提取算法等。本文重点研究了校园BBS舆情监测中的信息分析和舆情服务模块,主要有：1、改良的观点倾向识别技术在观点倾向识别技术研究中,分析了传统的识别技术,并研究了基于HOWNET词库的观点倾向性识别,针对回帖一般字数较少的特点还对回帖采用了计算否定词数量的观点倾向性识别算法改进。2、基于k-means算法的校园BBS聚类算法通过计算两个帖子关键词间的余弦相似度,可以得到两个帖子之间的距离(相似度),通过k-means算法将各个帖子按各自的距离进行聚类,能够得到较好的效果。3、基于关联规则自动更新的知识库匹配下的敏感帖监测技术敏感帖识别一般是基于知识规则库进行匹配,知识规则库中敏感词汇的更新是影响其性能的主要因素之一。当发现一个帖子是敏感帖后,本文还对这个帖子的其它关键字进行记录与分析,通过关联规则算法,将能够满足最小支持度阀值和最小置信度阀值的关键字收纳入知识规则库,实现了知识规则库的自动更新,经实验证明此方法支持下的敏感帖识别具有较高的准确率,并且能够随着识别次数的增加而提高识别准确率。4、热点话题识别与预测技术本文重点研究了热点话题识别与预测技术。其中在话题识别中针对校园BBS的特点提出了基于话题发帖量、话题浏览数、话题回复数以及话题精华帖量四个参数综合计算的话题识别方法。在热点话题预测中,研究了基于贝叶斯网络的热点话题预测技术,通过计算某个话题成为热点话题的可能性,成为普通话题的可能性以及成为冷门话题的可能性来预测该话题的舆情走向。通过这些研究,能够提高校园BBS舆情监测的效率,为校园BBS舆情监测提供技术支持。
[Abstract]:With the development of Internet, campus BBS has become one of the main platforms for college students to communicate online. Through the monitoring of campus BBS public opinion, we can collect data from campus BBS and analyze and process the data.According to the results of the analysis of accurate campus public opinion state control, campus public opinion discovery and early warning and other related operations.This paper analyzes the overall organizational framework of campus BBS public opinion monitoring. The campus BBS public opinion analysis system is divided into four modules: forum data acquisition module, data preprocessing module, public opinion information analysis module and public opinion monitoring and service module.This paper introduces some common techniques used in these four modules, mainly focused crawler algorithm and improved technology based on campus BBS, post denoising technology for BBS logical structure analysis, and Chinese word segmentation technology.Stop word technology, text classification method, keyword extraction algorithm and so on.This paper focuses on the research of information analysis and public opinion service module in campus BBS public opinion monitoring.And the view orientation recognition based on HOWNET thesaurus is studied.In view of the small number of words in general, the paper also uses an improved algorithm of attitude orientation recognition to calculate the number of negative words. The campus BBS clustering algorithm based on k-means algorithm calculates the cosine similarity between the keywords of two posts.We can get the distance between two posts (similarity degree), and cluster each post according to their own distance by k-means algorithm.It can get a good result. 3. Sensitive post recognition is based on knowledge rule base, which is based on the knowledge base matching, which is based on automatic updating of association rules.The update of sensitive words in the knowledge rule base is one of the main factors affecting its performance.When a post is found to be sensitive, this paper also records and analyzes the other keywords of this post. By using association rule algorithm, the key words that can satisfy the minimum support threshold and the minimum confidence threshold are incorporated into the knowledge rule base.The automatic updating of knowledge rule base is realized. It is proved by experiments that the sensitive post recognition based on this method has high accuracy.And can improve the recognition accuracy with the increase of recognition times. The hot topic recognition and prediction technology this paper focuses on the hot topic recognition and prediction technology.According to the characteristics of campus BBS, a topic recognition method based on the four parameters of topic posting quantity, topic browsing number, topic reply number and topic essence post quantity is put forward in topic recognition.In the prediction of hot topics, the technology of hot topic prediction based on Bayesian network is studied, and the possibility of a topic becoming a hot topic is calculated.The possibility of becoming an ordinary topic and the possibility of becoming a cold topic to predict the trend of public opinion on that topic.These studies can improve the efficiency of campus BBS public opinion monitoring and provide technical support for campus BBS public opinion monitoring.
【学位授予单位】：福州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09;TP391.1

【相似文献】