本文選題：精準(zhǔn)廣告投放　切入點(diǎn)：貝葉斯分類　出處：《華中師范大學(xué)》2013年碩士論文

【摘要】：隨著網(wǎng)絡(luò)技術(shù)的飛速發(fā)展,互聯(lián)網(wǎng)廣告成為互聯(lián)網(wǎng)企業(yè)最重要的盈利手段之一。越來越多的企業(yè)和機(jī)構(gòu)開始研究互聯(lián)網(wǎng)廣告平臺,與此同時,很多企業(yè)也慢慢地開始從傳統(tǒng)媒體廣告投放轉(zhuǎn)向互聯(lián)網(wǎng)廣告投放。然而,互聯(lián)網(wǎng)廣告投放的隨意性和泛濫性讓網(wǎng)民深受其煩,不僅網(wǎng)絡(luò)廣告的投放得不到預(yù)期的效果,而且網(wǎng)站點(diǎn)擊率也隨之下降。針對這種情況,互聯(lián)網(wǎng)廣告的精準(zhǔn)投放給互聯(lián)網(wǎng)廣告市場帶來了無限生機(jī)。精準(zhǔn)廣告投放即針對用戶的個性化向其投放感興趣的廣告,同時真正滿足用戶對產(chǎn)品需求的信息。 目前互聯(lián)網(wǎng)廣告系統(tǒng)中,要做到精準(zhǔn)投放主要有三種方式：常見的定向型,主要是針對地理位置、投放時間段等單個屬性或者組合屬性進(jìn)行投放；另一種是基于內(nèi)容的投放方式,這種廣告投放系統(tǒng)主要包括提取網(wǎng)頁主題詞、提取廣告文本主題詞,計(jì)算它們之間的相關(guān)性,然后進(jìn)行廣告的投放。而基于用戶行為特征的精準(zhǔn)廣告投放系統(tǒng)主要是在提取到用戶的行為特征數(shù)據(jù)之后,深入挖掘用戶的特征數(shù)據(jù),然后采用合適的分類算法對用戶分類,進(jìn)而針對用戶的特征投放廣告。 本文通過對互聯(lián)網(wǎng)廣告交易模式的進(jìn)一步分析,實(shí)現(xiàn)了一個互聯(lián)網(wǎng)廣告需求方平臺即DSP (Demand Side Platform)原型系統(tǒng),該系統(tǒng)通過與互聯(lián)網(wǎng)廣告交易平臺的對接,主要幫助廣告主參與到廣告的競拍中,并且綜合用戶信息、廣告信息等各種信息計(jì)算最佳待投放的廣告,從而實(shí)現(xiàn)廣告的精準(zhǔn)投放。 在用分類算法對用戶的特征分類時,常見的分類算法有神經(jīng)網(wǎng)絡(luò)分類算法、決策樹分類算法及貝葉斯分類算法等,但每種算法都有自己的優(yōu)缺點(diǎn),通過對比分析,選擇貝葉斯算法作為用戶特征分類算法。同時,考慮到每個屬性對類屬性不同的影響程度,運(yùn)用信息論的相關(guān)知識,設(shè)計(jì)出改進(jìn)的貝葉斯算法,經(jīng)過試驗(yàn)對比,改進(jìn)的貝葉斯算法比樸素貝葉斯算法的算法分類準(zhǔn)確率更高。
[Abstract]:With the rapid development of network technology, Internet advertising has become one of the most important means of profit for Internet enterprises. More and more enterprises and institutions begin to study Internet advertising platforms, at the same time, Many enterprises have also slowly begun to shift from traditional media advertising to Internet advertising. However, the randomness and flooding of Internet advertising have upset netizens deeply, not only that the online advertising cannot achieve the desired results, In response to this situation, the precise delivery of Internet advertising has brought infinite vitality to the Internet advertising market. At the same time, truly meet the needs of users for product information. In the current Internet advertising system, there are three main ways to achieve accurate delivery: the common directional type, mainly aimed at the geographical location, time periods and other single attributes or combination attributes; The other is content-based delivery. This advertising system mainly includes extracting the theme words of the web page, extracting the theme words of the advertising text, calculating the correlation between them. The accurate advertising system based on the user behavior features is mainly to extract the user behavior feature data, and then use the appropriate classification algorithm to classify the user. And then to the characteristics of the user advertising. In this paper, a prototype system of DSP demand Side platform is implemented through the further analysis of the mode of Internet advertising transaction, and the system is connected with the Internet advertising trading platform. It mainly helps advertisers to participate in the bidding of advertisements, and synthesizes all kinds of information, such as user information, advertising information and other information to calculate the best ads to be placed, so as to achieve the accurate placement of advertisements. When classifying users' features by using classification algorithms, the common classification algorithms include neural network classification algorithm, decision tree classification algorithm and Bayesian classification algorithm, but each algorithm has its own advantages and disadvantages. The Bayesian algorithm is chosen as the classification algorithm of user characteristics. At the same time, considering the different influence of each attribute on the class attribute, using the relevant knowledge of information theory, the improved Bayesian algorithm is designed and compared through experiments. The improved Bayesian algorithm is more accurate than the naive Bayesian algorithm.
【學(xué)位授予單位】：華中師范大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2013
【分類號】：TP393.09;TP311.52

【參考文獻(xiàn)】

相關(guān)期刊論文 前7條

1 楊暖暖;;論媒體創(chuàng)意與廣告形式創(chuàng)新[J];東南傳播;2010年01期

2 倪巍偉;陳耿;陸介平;吳英杰;孫志揮;;基于局部信息熵的加權(quán)子空間離群點(diǎn)檢測算法[J];計(jì)算機(jī)研究與發(fā)展;2008年07期

3 楊曉帆;陳廷槐;;人工神經(jīng)網(wǎng)絡(luò)固有的優(yōu)點(diǎn)和缺點(diǎn)[J];計(jì)算機(jī)科學(xué);1994年02期

4 周雪忠;吳朝暉;;文本知識發(fā)現(xiàn):基于信息抽取的文本挖掘[J];計(jì)算機(jī)科學(xué);2003年01期

5 丁春榮;李龍澍;楊寶華;;基于粗糙集的決策樹構(gòu)造算法[J];計(jì)算機(jī)工程;2010年11期

6 文炯;;搜索引擎之競價(jià)排名研究[J];江西圖書館學(xué)刊;2006年01期

7 季桂樹;陳沛玲;宋航;;決策樹分類算法研究綜述[J];科技廣場;2007年01期

相關(guān)碩士學(xué)位論文 前1條

1 張保華;數(shù)據(jù)挖掘技術(shù)的研究及在圖書借閱系統(tǒng)中的應(yīng)用[D];南京理工大學(xué);2008年

，

本文編號：1680340