基于用户标签的在线视频广告系统的研究
发布时间:2018-03-26 13:19
本文选题:在线视频广告 切入点:内容相关 出处:《华中科技大学》2012年硕士论文
【摘要】:随着在线视频行业的发展,视频行业的广告收入也一直保持着良好的增长势头。目前在线视频广告系统由于广告投放内容随机导致广告与视频内容的相关性较差、广告投放时机固定导致用户体验差,这两者极大的影响了广告的投放效果。同时现有的在线视频广告系统采用C/S模式进行广告分发,也给广告源服务器带来了较大的压力和带宽。 基于用户标签的在线视频广告系统CMAd能够有效提高广告的投放效果,并且降低广告源服务器在广告分发过程中的压力。CMAd从用户标签入手,从视频所有用户标签中提取出视频热点及其描述,根据视频热点描述选取出与之最相似的广告进行投放。由于用户标签较为准确的反应了视频当前播放内容,,因此可以有效提升广告与视频内容之间的相关性。CMAd设计了一种基于用户对视频关注度的广告投放时机选择策略,该策略通过找出视频热点区域中用户对视频关注度下降幅度最大的区间来进行广告投放,既能发挥在视频热点区域投放广告展示次数多、广告与视频内容相关性高的优势,又使得广告投放尽可能小的影响用户观看体验。CMAd采用无结构的拓扑组织方式,根据节点与邻居节点之间的缓存重叠度和网络延时动态更新节点的拓扑信息,最后利用节点的邻居节点辅助服务该节点的广告数据请求。 根据系统实际部署日志数据,在CMAd系统中,投放的广告与当前播放的视频内容相似度达到75%以上,用户关闭广告的概率低于10%,广告的点击率达到0.73%,广告源服务器带宽消耗相比采用C/S模式进行广告分发下降了70%。
[Abstract]:With the development of the online video industry, advertising revenue in the video industry has also maintained a good growth momentum. At present, the online video advertising system has a poor correlation between advertising and video content due to the random advertising content. Fixed advertising timing leads to poor user experience, which greatly affects the effectiveness of advertising. At the same time, the existing online video advertising system uses C / S mode for advertising distribution. Also bring greater pressure and bandwidth to the advertising source server. CMAd, an online video advertising system based on user label, can effectively improve the effect of advertising, and reduce the pressure of advertising source server in the process of advertising distribution. The video hotspots and their descriptions are extracted from all the user tags of the video, and the most similar advertisements are selected according to the description of the video hotspots. Because the user tags accurately reflect the current content of the video, Therefore, it can effectively enhance the correlation between advertising and video content. CMAd designs an advertising timing selection strategy based on users' attention to video. By finding out the region in which users have the biggest decrease in video attention, the strategy can not only play the advantages of more times of display and high correlation between advertising and video content, but also in the region of video hotspots. In addition, it makes the advertising as small as possible to affect the user viewing experience. CMAd adopts an unstructured topology organization mode, dynamically updates the topology information of nodes according to the cache overlap between nodes and neighbor nodes and network delay. Finally, the neighbor node of the node is used to assist the advertisement data request of the node. According to the actual deployment log data of the system, in the CMAd system, the similarity between the advertisement and the current video content is more than 75%. The probability of the user closing the advertisement is less than 10, the click rate of the advertisement reaches 0.73, and the bandwidth consumption of the advertising source server is 70% lower than that of using the C / S mode to distribute the advertisement.
【学位授予单位】:华中科技大学
【学位级别】:硕士
【学位授予年份】:2012
【分类号】:TP393.02
【参考文献】
相关期刊论文 前2条
1 俞淑平;陈刚;;一种高效的行为定向广告投放算法[J];计算机应用与软件;2011年01期
2 高媛;刘大中;;中文文本分类方法比较研究[J];科技信息(科学教研);2008年02期
相关硕士学位论文 前1条
1 俞淑平;网络定向广告投放算法研究[D];浙江大学;2010年
本文编号:1668016
本文链接:https://www.wllwen.com/wenyilunwen/guanggaoshejilunwen/1668016.html