基于LDA的行为定向广告投放算法研究

发布时间：2018-05-05 10:47

本文选题：行为定向广告 + 用户行为特征分析　；参考：《辽宁大学》2014年硕士论文

【摘要】：随着电子商务的发展，网络媒体广告成为网站主要的收入来源来源之一，层出不穷的网络广告投放技术导致了广告投放的无目的性，投放的广告很大程度上是用户不感兴趣的，这不仅不会引起用户的兴趣而去点击广告，还会招致用户的厌烦。针对这种情况，出现了精准广告投放这一概念。网络广告发展的精准化趋势催生了行为定向网络广告技术，这种广告模式具有精准性、及时性、高效性，从而备受人们的关注。行为定向广告的原理是挖掘用户浏览网页记录中包含的用户行为特征来投放广告，根据用户的行为特征，，广告主可以更精确地对目标用户进行广告投放,用最少的投资换取最高的利润。由于定向广告具有覆盖面广、精准度高的特点,被各大互联网公司和精准化广告服务商所重视。本文首先介绍了行为定向网络广告研究的背景和意义，行为定向广告的特点和投放过程中的难点，引出了本文的研究内容和创新点，之后介绍了广告投放精准化技术的现状。论述了网页预处理的过程，文档聚类常用的方法和常用的文档统计模型。然后针对行为定向广告，提出了一种基于LDA的行为定向广告算法的研究。该算法首先根据用户的历史访问记录，引入了追踪窗口的概念,根据网页访问时间的先后顺序进行排列,建立时变用户行为模型利用改进的k-means算法来计算用户的长期行为和短期行为；然后由时变用户行为模型结合LDA主题模型，提出了一种高效的行为定向广告投放算法LDA_M。最后根据LDA_M算法得到的最能代表用户行为的几个关键词与广告关键词进行匹配，选出和用户行为特征最相关的几个广告进行投放。该算法解决了传统广告投放算法不能分析语义和LDA模型广告投放算法不能很好的挖掘用户行为变化的缺点。实验表明，该算法能很好的区分用户的长期行为和短期行为，精确地挖掘用户的行为特征，投放与用户行为特征相关的广告，提高了广告投放的精确度。
[Abstract]:With the development of electronic commerce, the network media advertisement has become one of the main revenue sources of the website. The endless network advertising technology has led to the aimlessness of the advertisement, and the advertisement placed is to a large extent uninteresting to the users. This will not only not arouse the interest of users to click on advertising, but will cause users to be bored. In response to this situation, the concept of precision advertising has emerged. The trend of precision in the development of online advertising has given birth to behavioral orientation online advertising technology, this advertising model has precision, timeliness and efficiency, so people pay close attention to it. The principle of behavior oriented advertising is to excavate the user behavior characteristics contained in the user browsing web pages to place advertisements. According to the behavior characteristics of users, advertisers can more precisely place advertisements on the target users. The least investment in exchange for the highest profit. Because of its wide coverage and high accuracy, targeted advertising has been paid attention to by major Internet companies and precision advertising service providers. This paper first introduces the background and significance of the research on behavioral orientation online advertising, the characteristics of behavioral targeted advertising and the difficulties in the delivery process, leads to the research content and innovation of this paper, and then introduces the present situation of the precision advertising technology. This paper discusses the process of web page preprocessing, the common methods of document clustering and the commonly used document statistical model. Then, a behavioral orientation advertising algorithm based on LDA is proposed. The algorithm firstly introduces the concept of tracking window according to the user's historical access record, and arranges it according to the order of page access time. A time-varying user behavior model is established to calculate the long-term and short-term behaviors of users using an improved k-means algorithm, and then an efficient behavioral targeting advertising algorithm LDAS is proposed by combining the time-varying user behavior model with the LDA topic model. Finally, according to the LDA_M algorithm, the most representative of the user behavior of several keywords and advertising keyword matching, select the most relevant to the user behavior characteristics of several ads to put in. This algorithm solves the shortcomings that the traditional advertising algorithm can not analyze the semantics and the LDA model advertising algorithm can not well mining user behavior changes. Experiments show that the algorithm can distinguish the long-term behavior and short-term behavior of users accurately mining the behavior characteristics of users and putting in advertisements related to user behavior characteristics and improves the accuracy of advertising.
【学位授予单位】：辽宁大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092

【参考文献】