基于微博用户兴趣模型的个性化广告推荐研究

发布时间：2019-06-17 12:34

【摘要】：随着互联网技术以及信息传播技术的飞速发展,基于web2.0平台的微博等开放互联网社交服务模式越来越流行。在微博平台中,人人都像媒体一样可以自由发表感受和见解。近年来,基于微博的数据挖掘相关研究越来越多,本文通过构建微博用户兴趣模型,针对用户在微博平台发布的海量数据,挖掘能揭示用户兴趣点的关键主题词,并根据挖掘结果进一步深入探讨了如何实现个性化的广告推荐,从而帮助广告主们降低广告成本,提升广告的投放效果。本文对如何利用微博数据对用户兴趣进行分析,以及实现个性化广告推荐的方法和形式进行了研究和探索。与该领域已有的研究工作相比,本文主要有以下几点不同：首先,对不同的主题模型进行分析,比较了TwitterRank、Author-Topic和TwitterLDA三种主题模型在构建微博用户兴趣模型方面的性能,结合本文的研究内容,选择采用TwitterLDA模型进行新浪微博用户的兴趣识别。其次,将目前已有的改进后的LDA算法应用于微博用户主题词的挖掘,通过分析主题结构(topic structure)里的后验概率,来找出了能够表达主题含义的短语。改进后的算法既能保留传统LDA模型调换词序对主题挖掘结果没有影响的特点,同时还能使算法变得更高效,并获得了能表示主题含义的n-gram短语。最后,提出在微博个性化广告推荐的各种广告形式中融合故事型广告的创新模式并设计了以新浪微博普通用户为例的实证调研。最终通过对调研结果进行分析,验证了论文中使用的主题模型在普通微博用户中进行兴趣挖掘的可行性及有效性,并简单地就故事型广告的创新形式接纳度和兴趣模型的有效性进行了调研评估。通过本文的研究,可以发现,微博用户的行为和兴趣之间有很强的关联性,尤其是发布行为、转发行为和评论行为这三种主要行为。基于微博用户兴趣模型的个性化广告推荐研究能够分析微博用户的兴趣并进行精准的广告投放,降低广告成本,提高广告收益,带来更好的经济及社会效益。
[Abstract]:With the rapid development of Internet technology and information communication technology, open Internet social service models such as Weibo based on web2.0 platform are becoming more and more popular. In the Weibo platform, everyone is as free to express their feelings and opinions as the media. In recent years, there are more and more research on data mining based on Weibo. This paper constructs Weibo user interest model, mining the key subject words that can reveal the points of interest of users, and further discusses how to realize personalized advertising recommendation according to the mining results, so as to help advertisers reduce the cost of advertising and improve the effect of advertising. This paper studies and explores how to use Weibo data to analyze user interest and how to realize personalized advertising recommendation. Compared with the existing research work in this field, this paper mainly has the following differences: firstly, the different topic models are analyzed, and the performance of TwitterRank,Author-Topic and TwitterLDA in building Weibo user interest model is compared. combined with the research content of this paper, the TwitterLDA model is selected to identify the interest of Sina Weibo users. Secondly, the improved LDA algorithm is applied to the mining of topic words of Weibo users. By analyzing the posterior probability in the topic structure (topic structure), the phrases that can express the meaning of the topic are found out. The improved algorithm can not only preserve the characteristic that the traditional LDA model changing word order has no effect on the topic mining results, but also make the algorithm more efficient, and obtain the n-gram phrase which can express the meaning of the topic. Finally, this paper puts forward the innovative mode of integrating story-based advertising into various advertising forms recommended by Weibo personalized advertising, and designs an empirical investigation with Sina Weibo ordinary users as an example. Finally, through the analysis of the research results, the feasibility and effectiveness of the topic model used in this paper in interest mining among ordinary Weibo users are verified, and the innovative form acceptance of story advertising and the effectiveness of interest model are simply investigated and evaluated. Through the study of this paper, we can find that there is a strong correlation between the behavior and interest of Weibo users, especially the three main behaviors: publishing behavior, forwarding behavior and comment behavior. The personalized advertising recommendation research based on Weibo user interest model can analyze the interests of Weibo users and carry out accurate advertising, reduce advertising costs, improve advertising revenue, and bring better economic and social benefits.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：G358;F713.8

【参考文献】