一种分布式广告投放引擎的设计与实现

发布时间：2018-07-10 02:31

本文选题：定向广告 + 搜索引擎　；参考：《北京邮电大学》2012年硕士论文

【摘要】：随着互联网广告业的快速发展,定向广告作为一种新兴的网络广告模式也随之迅速发展,这种广告模式以其精准、及时、高效的特点备受人们的关注。定向广告是一种投放在网页上、与网页内容或者用户自身的行为特征相关的广告投放模式,按定向模式的不同可以分为内容定向广告和行为定向广告。与传统平面媒体、电视媒体相对而言,互联网的丰富交互手段使得互联网广告具备了传统媒体无法比拟的广告投放优化空间。PC终端丰富的多媒体交互能力、状态保持能力为广告定向投放行为提供了良好的平台。定向广告的基础是垂搜索引擎。垂直搜索引擎,是针对某一个行业的专业搜索引擎,是搜索引擎的细分和延伸,它针对某一特定领域、某一特定人群或某一特定需求提供、有一定价值的信息和相关服务。而广告垂直搜索引擎则是针对竞价广告行业的专业搜索引擎,它以结构化广告信息为搜索内容,向查询商品的用户提供相关广告的信息。垂直搜索引擎与通用网页搜索引擎有很大的不同,它和普通的网页搜索引擎的最大区别是对被查信息进行了结构化信息抽取,也就是将网页的非结构化数据变成特定的结构化信息数据,网贞搜索是以网页为最小单位,而垂直搜索是以结构化数据为最小单位,然后将这些数据存储到数据库,进行进一步的加工处理,如：去重、分类等,最后分词、索引再以搜索的方式满足用户的需求。广告垂直搜索引擎通要而对海量数据,引擎库中通常会存储几千万乃至上亿条的广告信息数据。在索引量和搜索量大到一定程度的时候,索引查询、更新的效率会逐渐降低,服务器的压力逐渐升高,整个搜索引擎的利用率可以说是越来越低了,并且随着海量数据存储带来的困难,设计一个良好的分布式垂直搜索引擎将成为一个垂直搜索引擎能否面向未来发展的关键因素。如何向用户推送最相关的广告,以及如何解决每天数以亿计的广告投放的问题是本文研究的重点。本文采用了文本内容匹配、长期用户行为分析、短期用户行为分析、关联推荐、融合模式等多种方法来为用户推送其最喜爱的广告,并通过多节点分布式计算体系,解决了超大规模并发下的实时请求处理问题。利用多层次数据处理架构解决了海量数据的汇总统计需求,实现出了一种在数据量、访问量增加的情况下可扩展的一种服务架构。
[Abstract]:With the rapid development of Internet advertising industry, targeted advertising as a new network advertising model has also developed rapidly. This advertising model has attracted people's attention for its accurate, timely and efficient characteristics. Targeted advertising is a kind of advertising mode which is placed on the web page and related to the content of the web page or the behavior characteristics of the user itself. It can be divided into content oriented advertising and behavioral targeted advertising according to the different orientation mode. Compared with the traditional print media and TV media, the rich interactive means of the Internet make the Internet advertisement have the multimedia interaction ability which the traditional media can not compare with the optimization space of advertising placement. PC terminal is rich in multimedia interaction. State-keeping ability provides a good platform for advertising targeting. The basis of targeted advertising is the vertical search engine. Vertical search engine is a professional search engine for a certain industry. It is the subdivision and extension of a search engine. It aims at a specific field, a specific population or a specific need to provide information and related services of a certain value. The vertical advertising search engine is a professional search engine for the bidding advertising industry. It uses structured advertising information as the search content to provide relevant advertising information to the users who query the products. The vertical search engine is very different from the general web search engine. The biggest difference between vertical search engine and common web search engine is the structured information extraction. In other words, the unstructured data of a web page is changed into a specific structured information data. Web search takes the web page as the smallest unit, while the vertical search takes the structured data as the minimum unit, and then stores the data into the database. Further processing, such as removing heavy, classifying, finally partitioning, indexing and searching to meet the needs of users. Advertising vertical search engines usually store tens of millions and even hundreds of millions of ad information data for mass data. When the number of indexes and searches is large to a certain extent, the efficiency of index queries and updates will gradually decrease, the pressure on the server will gradually increase, and the utilization rate of the whole search engine can be said to be getting lower and lower. With the difficulties of mass data storage, the design of a good distributed vertical search engine will be a key factor for the future development of a vertical search engine. How to push the most relevant ads to users and how to solve the problem of hundreds of millions of ads every day is the focus of this paper. In this paper, text content matching, long term user behavior analysis, short term user behavior analysis, association recommendation, fusion mode and other methods are used to push their favorite advertisements for users, and multi-node distributed computing system is adopted. It solves the problem of real-time request processing under super large scale concurrency. The multilevel data processing architecture is used to solve the statistical requirement of mass data, and an extensible service architecture is implemented with the increase of data volume and traffic.
【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】：TP311.52

【相似文献】