基于微博客话题的热点预测及传播溯源

发布时间：2018-06-09 17:40

本文选题：微博话题 + 热点预测　；参考：《北京邮电大学》2014年硕士论文

【摘要】：微博作为一种新型的自媒体平台,以其信息传播速度快、信息源多、用户参与度高而得到迅速发展。同时,微博的这种传播特性也导致信息的庞大而杂乱,信息的真伪性和话题的热度等成为人们关注的重点。本文根据微博分析中的热门应用,研究了微博热点话题预测与话题传播途径分析方面的相关技术。本文采用话题源头假设模型研究新浪微博的热点话题预测。该方法将微博用户群抽象为话题源头,并根据热门话题和非热门话题由话题源头产生的假设,将热点话题预测问题转化为话题的分类问题。同时,论文实现并改进了一种基于微博话题关注度的时间序列结构化学习的方法,对热门话题进行预测。实验结果证明该方法能以90%的准确率分出热门话题,其中77.8%的话题的预测点提前于其上官方微博热门话题榜的时间,提前的平均时间为1.54小时。另外,本文根据单条微博与微博话题的转发评论信息,重构了话题传播途径,并将其可视化呈现。在此基础上,本文亦对话题传播途径中参与用户的影响力做了分析,实现了一种话题传播中关键用户的计算方法,可以有效的进行微博话题溯源分析。
[Abstract]:As a new type of self-media platform, Weibo has been developed rapidly because of its high speed of information dissemination, more information sources and high user participation. At the same time, the spread of Weibo also leads to the huge and messy information, the authenticity of information and the heat of topics become the focus of attention. According to the popular application of Weibo, this paper studies the techniques of Weibo hot topic prediction and topic propagation approach analysis. This paper uses the topic source hypothesis model to study the hot topic prediction of Sina Weibo. This method abstracts the Weibo user group as the topic source and transforms the hot topic prediction problem into the topic classification problem according to the hypothesis that hot topic and non-hot topic are generated by topic source. At the same time, this paper implements and improves a structured learning method of time series based on Weibo topic concern to predict hot topics. The experimental results show that the method can distinguish hot topics with 90% accuracy, 77.8% of the topics are predicted earlier than the official Weibo hot topics list, and the average advance time is 1.54 hours. According to the single Weibo and Weibo topic forwarding comments, this paper reconstructs the topic communication approach and visualizes it. On this basis, this paper also analyzes the influence of the users involved in topic communication, and realizes a calculation method of key users in topic communication, which can effectively analyze the origin of Weibo topics.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092

【参考文献】