基于视频流的用户兴趣挖掘模型设计及仿真实现

发布时间：2018-03-22 19:01

本文选题：汇聚设备　切入点：视频流　出处：《电子科技大学》2012年硕士论文　论文类型：学位论文

【摘要】：在信息时代的今天，互联网应用已渗透到各行各业乃至在日常生活中。在这种趋势下，，电子商务高速发展，而在网络上投放广告已经成为电子商务重要的营销方式。然而目前，粗放的广告投放并不能得到让人满意的效果，甚至造成负面效果，因此针对用户的兴趣所在实施精准广告的定向投放是网络广告大势所趋。另一方面，随着网络带宽跃升，网络传输速度大幅提高，传输质量有了相当的保障，视频流技术得到迅速发展和广泛运用。作为传统电信领域的龙头，运营商为了改变传统的单纯依靠带宽收费的盈利模式，运营商已经进行了很多新业务的尝试，其中用户定向广告投放已经进入应用推广阶段。用户兴趣挖掘是定向广告的基础和前提。针对文字Web的兴趣挖掘技术已经相当成熟，而基于视频流的相关技术却还在探索之中。论文首先介绍了目前的视频流信息挖掘技术中的文本和行为信息的获取及表示技术。然后，论文讨论了在汇聚设备针对视频流而挖掘用户兴趣需要而且可以获取哪些信息，如何得到这些信息。并且结合现有的技术，设计出针对视频流领域特有的用户兴趣模型表示方法标准分类树SCT(StandardCategoryTree)。标准分类树方法是改进的层次性本体论方法。使用这种方法判断用户兴趣，首先需要建立树状层次的视频领域标准分类，每个分类关联特征词的训练集和该分类对应的人群概率。特征词训练集用于对用户的观看视频进行分类，人群概率用于对用户所属人群的概率判断。根据用户观看所观看视频的类型判断用户的人群概率，然后借助人群概率计算出用户感兴趣的商品分类和兴趣度。接着，论文设计并且实现了实验系统来验证标准分类树方法在用户兴趣判断方面的可行性和准确性。实验的主要工作是建立标准分类树，每个分类建立特征词训练集，输入用户观看视频的历史信息，包括视频名称、视频分类、视频描述、演员、地区、视频时长、用户观看日期和用户观看时长，输出该用户对各商品分类的兴趣度并排名。最后，论文分析和总结了将标准分类树方法应用于视频流领域的用户兴趣挖掘的优点和缺点。并且对相关技术的发展进行了展望。
[Abstract]:In the information age, Internet application has penetrated into all kinds of industries and even in daily life. Under this trend, E-commerce has developed rapidly, and placing advertisements on the Internet has become an important marketing method of E-commerce. Extensive advertising can not get satisfactory results, or even cause negative effects, so it is the trend of network advertising to target the interests of users to carry out targeted advertising. On the other hand, with the rise of the network bandwidth, the transmission speed of the network has been greatly improved, the transmission quality has been quite guaranteed, and the video streaming technology has been rapidly developed and widely used. In order to change the traditional profit mode which relies solely on bandwidth charges, operators have made a lot of new business attempts, among which targeted advertising has entered the stage of application promotion. User interest mining is the basis and premise of targeted advertising. Interest mining technology for text Web has been quite mature. However, the related technology based on video stream is still being explored. Firstly, this paper introduces the text and behavior information acquisition and representation technology in the current video stream information mining technology. Then, the paper discusses what information can be obtained and what information can be obtained by mining user interest in convergent devices for video streams, and combines with existing technologies. This paper designs a representation method of user interest model for video stream domain. The standard classification tree SCT / Standard Category tree is an improved hierarchical ontology method, which is used to judge user interest. First of all, it is necessary to establish the standard classification of video domain at the tree level, the training set of each classification association feature word and the corresponding crowd probability of the classification. The training set of feature words is used to classify the user's watching video. The crowd probability is used to judge the population probability of the user, and the group probability of the user is judged according to the type of video the user is watching, and then the classification and interest degree of the goods of interest to the user are calculated by the crowd probability. Then, the experiment system is designed and implemented to verify the feasibility and accuracy of the standard classification tree method in judging the user's interest. The main work of the experiment is to establish the standard classification tree, and each classification establishes the training set of feature words. Input the historical information of the user to watch the video, including the video name, video classification, video description, actor, region, video duration, user viewing date and user viewing time, output the user's interest in each item classification and rank. Finally, the paper analyzes and summarizes the advantages and disadvantages of applying the standard classification tree method to user interest mining in video streaming field, and prospects the development of related technologies.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP393.09

【参考文献】