基于云平台的大型视频网站广告投放系统的设计与实现

发布时间：2018-03-11 22:13

  本文选题：广告投放　切入点：海量数据　出处：《哈尔滨工业大学》2016年硕士论文　论文类型：学位论文

【摘要】：在线广告投放系统是随着在线广告市场和大数据技术的的快速发展和逐步完善而发展起来的,旨在帮助广告商和在线视频网站实现专业精准的贴片类、页面类广告定向投放,广告投放流程管理,广告运营数据统计等任务的系统软件。56网作为中国领先的视频分享网站,每天视频播放量超过千万次,公司搭建在线广告投放管理系统,目的在于利用云计算平台,通过对广告投放流程的严格把控,在保证广告定向精准投放的前提下,同时能够整合公司既有的销售、产品、技术团队,形成可靠多样的,可定制化的广告销售方案。该广告系统能够对广告投放方案进行有效的优化,并在投放的过程中自动调节投放参数,从技术层面确保了网站广告资源最大程度的有效利用。本课题通过对计算广告学、流程管理理论、相似度推荐算法、弹性计算和海量存储技术等相关理论和技术的探究,利用云计算平台搭建了一个支持日千万级用户访问的在线广告智能投放、投放流程管理、投放数据挖掘的在线广告投放系统,该广告系统能够对网站广告资源进行全程监控和控制,使不同连接设备、不同观看时间,不同地理位置的用户在观看网站视频时,都能够根据广告商制定的投放方案获得差异化的广告投放结果,使品牌在目标用户群中的曝光频率大幅提高。本系统的开发过程包含对投放系统硬件集群部署,广告投放策略优化,投放管理后台系统开发,海量数据分析与预测等方面,主要实现了以下几方面的功能:搭建了广告后端投放子系统和数据采集分析子系统,实现了对广告投放的控制和数据的采集;设计并开发了用于预订、管理、投放广告资源的流程管理子系统;设计并开发了合同管理、下单管理子系统,实现了针对用户画像的广告精准投放;设计并开发了投放数据监测与效果反馈子系统;设计并开发了广告系统的数据盘点子系统。通过对各个子系统的不同层次的测试,该广告投放系统满足了56网提出的功能和非功能性需求,并在此基础上实现了基于皮尔逊相关系数、Cosine相似度、Tanimoto系数等利用群体智慧协同过滤的广告推荐算法。
[Abstract]:With the rapid development and gradual improvement of online advertising market and big data technology, the online advertising delivery system aims to help advertisers and online video websites to achieve professional and accurate patch type, page type advertising targeted delivery. As a leading video sharing website in China, the system software .56. net for tasks such as advertising process management, advertising operation data statistics, and so on, has more than 10 million video playback times a day. The company builds an online advertising delivery management system. The purpose is to make use of cloud computing platform, by strictly controlling the advertising delivery process, while ensuring the targeted and accurate delivery of advertisements, and at the same time integrating the company's existing sales, products, and technical teams to form a reliable and diverse, The advertising system can effectively optimize the advertising delivery scheme and automatically adjust the delivery parameters in the process of launching. From the technical level to ensure the most effective use of website advertising resources. This topic through the computational advertising theory, process management theory, similarity recommendation algorithm, flexible computing and mass storage technology and other related theories and technologies. Using the cloud computing platform to build an online advertising delivery system that supports daily users to access the online advertising intelligent delivery, delivery process management, data mining, online advertising delivery system, The advertising system can monitor and control the advertising resources of the website in the whole process so that users with different connecting devices, different viewing times and different geographical locations can watch the website video. All of them can get different advertising results according to the advertising program, which can greatly increase the exposure frequency of the brand in the target user group. The development process of the system includes the deployment of the hardware cluster of the delivery system. The optimization of advertising strategy, the development of backstage system of launch management, the analysis and prediction of massive data, etc., are mainly realized in the following aspects: the back-end system of advertising and the subsystem of data collection and analysis are set up. It realizes the control of advertising and data collection, designs and develops a process management subsystem for booking, managing and placing advertising resources, designs and develops a contract management subsystem and an order management subsystem. In this paper, the accurate advertisement placement for user portrait is realized; the subsystem of data monitoring and effect feedback is designed and developed; the data inventory subsystem of advertising system is designed and developed. Through the test of different levels of each subsystem, The advertising delivery system meets the functional and non-functional requirements proposed by 56net, and based on this, an ad recommendation algorithm based on Pearson correlation coefficient and Cosine similarity and Tanimoto coefficient is implemented.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.52
，

本文编号：1600142