基于行为定向的精准广告投放系统的研究与实现
发布时间：2018-07-27 15:08
【摘要】：随着科技的进步和人们生活水平的提高,互联网用户大量增加,据IDC报告称2012年全球网民数量将突破19亿,中国网民数量将突破5亿。如此大的用户规模,使各类企业趋向于借助于网络广告推广自己的产品和服务。目前网络广告存在着无目标、爆炸式投放的缺点,无法满足企业的推广预期,关于广告精准投放的研究在深入进行。 目前网络广告精准化投放主要有三种方式,基于物理位置的分类广告、基于用户行为/网页内容的定向广告、基于手机终端的短信广告等。其中定向广告的覆盖面广、精准度高,被各大互联网公司和精准化广告服务商所重视。课题主要研究基于用户行为定向的精准化广告投放,并且进行了设计和实现。论文首先介绍一些背景知识和精准化投放技术的现状,之后介绍系统实现中的相关术语、技术、以及所使用的分类算法,重点介绍系统设计实现的过程,包括数据库设计以及各个功能模块的详细设计等。 在精准效果优化章节,提出了一种优化算法,即将用户的行为分为短期行为和长期行为,通过网络广告追踪窗口对用户浏览的网页进行追踪。根据用户上网行为的Recency,Intensity,Frequenty值及系统规定的权重,分别对用户长期行为和短期行为进行特征分析,挖掘出用户的短期偏好和长期偏好并计算出Raw Score。根据长期行为和短期行为的权重以及Raw Score,计算出用户各个主题的分值矩阵,并据此进行广告投放,实验数据显示,经过优化模块的处理,广告投放精准度有一定的提高,此模块是整个系统的创新点。 论文最后给出了全文的总结,提出系统中存在的不足情况、下步的工作重点,总结了本人在硕士研究生期间的工作和成果。
[Abstract]:With advances in technology and living standards, Internet users have increased dramatically, with the number of global netizens set to exceed 1.9 billion in 2012 and the number of Internet users in China to exceed 500m, according to the IDC report. With such a large scale of users, all kinds of enterprises tend to promote their products and services by means of network advertisement. At present, the network advertisement has the shortcoming of aimless and explosive delivery, which can not meet the enterprise's promotion expectation, and the research on the precise advertising is carried out in depth. At present, there are three main ways to precisely deliver online advertisements, such as classified ads based on physical location, targeted advertisements based on user behavior / web content, SMS ads based on mobile terminals, etc. One of the wide coverage of targeted advertising, high accuracy, by the major Internet companies and precision advertising services. This paper mainly studies the precision advertising based on user behavior orientation, and designs and implements it. This paper first introduces some background knowledge and the current situation of precision delivery technology, then introduces the relevant terminology, technology, classification algorithms used in the system implementation, and focuses on the process of system design and implementation. Including database design and the detailed design of each functional module. In the chapter of precision effect optimization, an optimization algorithm is proposed, which divides the user's behavior into short-term behavior and long-term behavior, and tracks the web pages viewed by users through the web advertising tracking window. According to the frequent values of users' online behaviors and the weights prescribed by the system, the characteristics of long-term and short-term behaviors of users are analyzed, and the short-term and long-term preferences of users are excavated and the Raw Scoreis calculated. According to the weight of long-term behavior and short-term behavior and Raw Score, this paper calculates the score matrix of each topic of user, and carries on the advertisement placement accordingly. The experimental data shows that the precision of advertisement delivery has been improved by the processing of optimization module. This module is the innovation of the whole system. Finally, the paper gives a summary of the full text, puts forward the deficiencies in the system, the next step of the focus of work, summed up my work and achievements in the period of master's graduate.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP393.09
本文编号：2148217
[Abstract]:With advances in technology and living standards, Internet users have increased dramatically, with the number of global netizens set to exceed 1.9 billion in 2012 and the number of Internet users in China to exceed 500m, according to the IDC report. With such a large scale of users, all kinds of enterprises tend to promote their products and services by means of network advertisement. At present, the network advertisement has the shortcoming of aimless and explosive delivery, which can not meet the enterprise's promotion expectation, and the research on the precise advertising is carried out in depth. At present, there are three main ways to precisely deliver online advertisements, such as classified ads based on physical location, targeted advertisements based on user behavior / web content, SMS ads based on mobile terminals, etc. One of the wide coverage of targeted advertising, high accuracy, by the major Internet companies and precision advertising services. This paper mainly studies the precision advertising based on user behavior orientation, and designs and implements it. This paper first introduces some background knowledge and the current situation of precision delivery technology, then introduces the relevant terminology, technology, classification algorithms used in the system implementation, and focuses on the process of system design and implementation. Including database design and the detailed design of each functional module. In the chapter of precision effect optimization, an optimization algorithm is proposed, which divides the user's behavior into short-term behavior and long-term behavior, and tracks the web pages viewed by users through the web advertising tracking window. According to the frequent values of users' online behaviors and the weights prescribed by the system, the characteristics of long-term and short-term behaviors of users are analyzed, and the short-term and long-term preferences of users are excavated and the Raw Scoreis calculated. According to the weight of long-term behavior and short-term behavior and Raw Score, this paper calculates the score matrix of each topic of user, and carries on the advertisement placement accordingly. The experimental data shows that the precision of advertisement delivery has been improved by the processing of optimization module. This module is the innovation of the whole system. Finally, the paper gives a summary of the full text, puts forward the deficiencies in the system, the next step of the focus of work, summed up my work and achievements in the period of master's graduate.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP393.09
【参考文献】
相关期刊论文 前4条
1 王中华;;网络广告模式现存问题及解决思路探讨[J];经济研究导刊;2010年01期
2 郑欣杰;朱程荣;熊齐邦;;基于MapReduce的分布式光线跟踪的设计与实现[J];计算机工程;2007年22期
3 刘磊;陈兴蜀;尹学渊;段意;吕昭;;基于特征加权朴素贝叶斯分类算法的网络用户识别[J];计算机应用;2011年12期
4 沈维梅;;网络精准广告的发展及困惑[J];新闻界;2010年01期
相关博士学位论文 前1条
1 胡静;机器学习及其神经网络分类器优化设计[D];合肥工业大学;2007年
相关硕士学位论文 前4条
1 王凯;MapReduce集群多用户作业调度方法的研究与实现[D];国防科学技术大学;2010年
2 杨立娟;手机广告产业发展及中国移动推广策略研究[D];北京邮电大学;2008年
3 张保华;数据挖掘技术的研究及在图书借阅系统中的应用[D];南京理工大学;2008年
4 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：2148217
本文链接：https://www.wllwen.com/wenyilunwen/guanggaoshejilunwen/2148217.html
