Font Size: a A A

The Design And Implementation Of Advertising Strategy And Anti-fraud System For Renren.com

Posted on:2013-11-10Degree:MasterType:Thesis
Country:ChinaCandidate:M ZhangFull Text:PDF
GTID:2248330371485167Subject:Software engineering
Abstract/Summary:PDF Full Text Request
Traditional advertising tend to face to the majority, blind and not accurate enough. Thefirst modern advertiser, John Wanamaker, complained more than100years ago:"I know halfof my money spent on advertising is wasted, but the trouble is I don’t know which half ".Internet is the media developing fastest in the history. It is accepted and favorite bypeople for the advantage of spreading information rapidly and wide coverage. SNS is theabbreviation of the Social Network Service or Social Network Site, according to the theory ofsix degrees of separation: one will be able to recognize a stranger by a chain of people notmore than six. Each person’s social circle can expand via the current circle of friends, andeventually became a large social network.With the rapid development of the Internet and the emerging of the SNS website, theonline advertising industry based on the Internet and SNS become a new field for profits.Computational Advertising comes from online advertising practices, and becomes thefoundation of science and technology for online advertising industry now. Thanks to thedevelopment of the Internet and Computational Advertising, John Wanamaker’s complainmentioned before has solved mainly now, the waste of advertisers’ money becomes smallerand smaller and the value of advertising can be measured more easily. The ads onRenRen.com relying on a large number of real users, so they put forward the slogan "The adson RenRen.com can be targeted to everyone accurately ",To achieve this aim, they analyze thevaluable information of user profile and users’ browsing actions.By analyzing the page view logs and ads click logs in advertising system, we found thatthere exists some abnormal behavior such as clicking ads twice or clicking ads immediatelyafter the impression or clicking ads after a long time. Some ads is viewed by one user butclicked by another. Similarly, Some ads are viewed by user from one IP address but clicked byuser from another IP address. Because the ads are charged for CPM(Cost per mille/Cost perThousand Impressions) or CPC(Cost Per Click), if we do not take appropriate anti-fraudmeasures to deal with the behavior mentioned before, advertiser’s return on investment willdecrease and the interests of advertiser will be injured. The online advertising fraud willimpact a negative effect on the whole online advertising industry for a long term.To the problems that ads targeting is not accurate enough and the ads budget is wasted,the paper describes the solution-application of computational advertising-advertisingplatform of RenRen.com, summarizes the recommendation algorithm of ads from severalaspect: ads recommendation policy, the context of ads, fit evaluation of user’s profile and ads etc.Several forms of on online advertising fraud including click or view ads by programs andscripts or click on ads by taking part in advertising fraud groups, My independent work isputting forward the design, implementation and improvement direction of a viable CPConline anti-fraud system. I make strategy of anti-fraud filter based on the analysis andstatistics of ads page view log and click log, then develop and test the system finally.Advertising strategy module is written by C++, and the CPC online anti-fraud system isdeveloped using JAVA language. I take advantage of Thrift as RPC framework that implementthe function of Remote Procedure Call between various of s ystems, Hadoop is used as adistributed computing platform which saving all ads page view logs and click log. Hiveframework is used for getting data from HDFS by Hive Query Language (SQL-likelanguage).
Keywords/Search Tags:Computational Advertising, ads targeting accurately, SNS ads, online ads anti-fraud
PDF Full Text Request
Related items