微博消息的多元可信度模型研究与应用

发布时间：2018-08-26 16:06

【摘要】：微博消息真伪混杂，，虚假消息传播速度快、影响范围广，对社会、企业或个人能造成极大的不良影响。人为判断消息的真伪在时效性和准确性上存在很大挑战，因而构建智能的、有效的识别虚假消息的机制,评估消息可信度成为一个非常迫切而重要的问题。本文以主流微博上的消息和用户为研究对象，结合自然语言处理、统计学习以及行为分析等理论构建模型。本文的研究工作主要体现在以下两个方面：第一，微博水军识别研究。微博用户在微博平台中起主导作用，识别微博水军是评估消息可信度的重要过程，目前识别微博水军多为人工识别，难以在大数据中达到理想效果。本文通过分析微博中普通用户和水军的属性和行为差异，定义识别水军的有效特征，对特征的准确性进行了实验性对比，利用概率图模型的思想，构建了自动识别微博水军的WGM模型。将国内外最具代表性的两个微博平台新浪微博和Twitter数据作为实验样本，进行大量对比实验，结果显示利用提出的特征可以有效的识别微博平台中的水军，且WGM模型比传统的机器学习方法更准确、有效。第二，微博谣言检测与消息可信度研究。评估微博消息可信度，对消息真伪做出判别是一个非常重要但困难的问题。本文从微博评论出发，定义支持性、内容相关性、置信度三个特征衡量评论在谣言与真实微博中的差异性，利用评论特征构建BPCM模型评估消息可信度，判别消息真伪。采用新浪微博真实数据对特征与模型进行详细对比分析，结果表明利用微博评论可以有效的评估消息可信度，判别消息真伪，特征对谣言与真实微博的评论区分度较好，BPCM模型可以有效地解决消息可信度的评估问题以及谣言检测问题。本课题的研究内容有助于国家舆情监控、企业营销分析以及领域事件分析和精准搜索。
[Abstract]:Weibo news is mixed, the false news spreads quickly, the influence scope is wide, can cause the great adverse influence to the society, the enterprise or the individual. There are great challenges in judging the validity and accuracy of messages, so it is an urgent and important problem to construct an intelligent and effective mechanism to identify false messages and evaluate the credibility of messages. In this paper, we take the messages and users on the mainstream Weibo as the research object, and combine the natural language processing, statistical learning and behavior analysis theory to construct the model. The research work of this paper is mainly reflected in the following two aspects: first, Weibo water army recognition research. Weibo users play a leading role in the Weibo platform, and recognizing Weibo Navy Army is an important process to evaluate the credibility of news. At present, the recognition of Weibo Navy Army is mostly human identification, so it is difficult to achieve the ideal effect in Weibo. By analyzing the differences in attributes and behaviors between ordinary users and navy troops in Weibo, this paper defines and identifies the effective features of the naval forces, makes an experimental comparison of the accuracy of the features, and makes use of the idea of probability map model. A WGM model for automatic identification of Weibo navy army is constructed. Taking the data from the two most representative Weibo platforms at home and abroad, Sina Weibo and Twitter, as experimental samples, a large number of comparative experiments have been carried out. The results show that the proposed features can be used to effectively identify the watermen in the Weibo platform. The WGM model is more accurate and effective than the traditional machine learning method. Second, Weibo rumour detection and information credibility research. It is a very important but difficult problem to assess the credibility of Weibo messages and to judge the authenticity of messages. Based on Weibo's comments, this paper defines three characteristics of support, content relevance and confidence to measure the differences between rumors and true Weibo, and constructs a BPCM model to evaluate the credibility of messages and judge whether the messages are true or false. Using the real data of Sina Weibo to compare and analyze the features and models in detail, the results show that the use of Weibo comments can effectively evaluate the credibility of the message and judge whether the message is true or false. The BPCM model can effectively solve the problem of message credibility evaluation and rumor detection. The research content of this topic is helpful to national public opinion monitoring, enterprise marketing analysis, field event analysis and accurate search.
【学位授予单位】：北京工商大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092

【参考文献】