【摘要】：在線廣告如今已經(jīng)成為一個擁有200億美元的產(chǎn)業(yè)。它在成為眾多網(wǎng)站提供商和一些廣告中介商的重要收益來源的同時,也為廣告商提供了新的媒體平臺。為了使得廣告中介商更有效地為終端用戶提供在線廣告,從而增加廣告的點擊量;我們必須首先了解廣告中介商網(wǎng)絡(luò)在網(wǎng)絡(luò)層面與內(nèi)容層面運行情況的不足,然后再進一步地進行改進。本論文首先測量了現(xiàn)有的三個主流在線廣告網(wǎng)絡(luò)平臺:Google、AOL/Akamai和Adblade,并量化了他們網(wǎng)絡(luò)層面和內(nèi)容層面的性能。在此基礎(chǔ)上,本論文從基于用戶地理位置和基于用戶行為的在線廣告投放機制這兩個方面來改進現(xiàn)有的在線廣告網(wǎng)絡(luò)。本論文的主要創(chuàng)新性成果包括以下四個方面: 1.本論文是學術(shù)領(lǐng)域中第一篇從網(wǎng)絡(luò)層面和內(nèi)容層面測量在線廣告網(wǎng)絡(luò)的文章。本論文設(shè)計并實現(xiàn)了一套從網(wǎng)絡(luò)層面和內(nèi)容層面審計在線廣告網(wǎng)絡(luò)的系統(tǒng)。該系統(tǒng)可以廣泛運用于任一廣告中介商網(wǎng)絡(luò)中,有效地監(jiān)控并幫助調(diào)控基于Web的廣告業(yè)。通過使用這個系統(tǒng),本論文對擁有不同設(shè)計理念的廣告中介商在網(wǎng)絡(luò)層面和內(nèi)容層面進行了分析,并證明了在線廣告網(wǎng)絡(luò)的底層設(shè)計與上層的網(wǎng)絡(luò)層面和內(nèi)容層面的服務(wù)有很大的關(guān)系。 2.本論文設(shè)計并實現(xiàn)了一套精確度逐步提高的、無需客戶端支持的IP地址位置定位系統(tǒng)。該系統(tǒng)把Web的數(shù)據(jù)挖掘和Internet網(wǎng)絡(luò)測量的定位方法有效結(jié)合在一起,從而避免了單獨使用任何一種方法所帶來的不精確因素。通過對學校IP數(shù)據(jù)集上的測試,本論文證實這套系統(tǒng)的中值誤差距離僅為690米。這比之前最好的IP地理定位系統(tǒng)的準確度提高了50倍。因此,當這套系統(tǒng)嵌入到基于用戶地理位置的在線廣告網(wǎng)絡(luò)中,它可以很大程度上改進在線廣告投放的有效性。 3.本論文設(shè)計并實現(xiàn)了一套不需分析用戶數(shù)據(jù)包有效載荷,可高準確度地恢復(fù)用戶網(wǎng)頁瀏覽記錄的系統(tǒng)。該系統(tǒng)把Web網(wǎng)站爬尋統(tǒng)計和Internet網(wǎng)絡(luò)流量分析方法有效結(jié)合,而不使用違反互聯(lián)網(wǎng)法律的深度包檢測技術(shù)。通過恢復(fù)大規(guī)模級別的網(wǎng)絡(luò)流中的用戶瀏覽記錄,本論文證實這套系統(tǒng)可以達到86%以上的恢復(fù)成功率,和5%以下的誤報率。因此,當這套系統(tǒng)嵌入到基于用戶行為的在線廣告網(wǎng)絡(luò)中,它可以從根本上改變在線廣告投放的合法性。 4.本論文設(shè)計并實現(xiàn)了一套根據(jù)用戶的瀏覽模式來準確自動地標識用戶在線興趣的系統(tǒng)。對于每一項用戶瀏覽記錄,該系統(tǒng)通過網(wǎng)站的域名或者在搜索引擎返回的結(jié)果上,采用關(guān)鍵字匹配的方法來實現(xiàn)興趣分類。通過對三個月包含4.5萬個用戶的DNS網(wǎng)絡(luò)流進行實驗,本論文證實該套系統(tǒng)的正確率可以86.44%。因此,當這套系統(tǒng)嵌入到基于用戶行為的在線廣告網(wǎng)絡(luò)中,它可以很大程度上改進在線廣告投放的有效性。
[Abstract]:Online advertising has become an industry of $20 billion. It not only becomes an important source of revenue for many website providers and some advertising agencies, but also provides a new media platform for advertisers. In order to make the advertising agency more effective to provide online advertising to end users, thus increasing the number of ads click; We must first understand the shortcomings of the advertising intermediary network at the network level and the content level, and then further improve it. In this paper, we first measure the three main online advertising network platforms: Google,AOL/Akamai and Adblade, and quantify their performance at the network level and content level. On this basis, this paper improves the existing online advertising network from two aspects: user geographical location and online advertising mechanism based on user behavior. The main innovative achievements of this paper include the following four aspects: 1. This paper is the first in the academic field to measure the online advertising network at the network level and content level. This paper designs and implements a set of online advertising network audit system from the network level and content level. The system can be widely used in any advertising agency network and can effectively monitor and control the advertising industry based on Web. Through the use of this system, this paper analyzes the advertising agencies with different design concepts at the network level and content level. It is proved that the bottom layer design of online advertising network is closely related to the service at the network and content level. 2. In this paper, we design and implement a IP address location system with improved precision and no client support. The system combines the data mining of Web and the positioning method of Internet network measurement effectively, thus avoiding the imprecise factors brought by using any method alone. The median error distance of this system is only 690m by testing the school IP data set. This is 50 times more accurate than the previous best IP geographic positioning system. Therefore, when the system is embedded in the online advertising network based on users' geographical location, it can greatly improve the effectiveness of online advertising. 3. In this paper, we design and implement a system that can recover the user's browsing record accurately without analyzing the payload of the user's packet. This system combines Web website crawling statistics with Internet network traffic analysis method effectively, and does not use the depth packet detection technology which violates the Internet law. By restoring user browsing records in large-scale network flows, this paper confirms that the system can achieve a recovery success rate of more than 86% and false alarm rate of less than 5%. Therefore, when the system is embedded in the online advertising network based on user behavior, it can fundamentally change the legitimacy of online advertising. 4. In this paper, we design and implement a system to identify the online interest of users accurately and automatically according to the user's browsing mode. For each user browsing record, the system uses keyword matching method to realize interest classification through the domain name of the website or the result returned by the search engine. Through the experiments on the DNS network flow with 45000 users in three months, this paper proves that the correct rate of the system can be 86.44. Therefore, when the system is embedded in the online advertising network based on user behavior, it can greatly improve the effectiveness of online advertising.
【學位授予單位】：電子科技大學
【學位級別】：博士
【學位授予年份】：2011
【分類號】：TP393.092

【共引文獻】

相關(guān)碩士學位論文 前1條

1 姚園園;Internet拓撲結(jié)構(gòu)研究及在IM網(wǎng)絡(luò)建模中的應(yīng)用[D];鄭州大學;2006年

，

本文編號：2302992