本文選題：行為定向廣告 + 用戶行為特征分析　； 參考：《遼寧大學》2014年碩士論文

【摘要】：隨著電子商務的發(fā)展，網(wǎng)絡媒體廣告成為網(wǎng)站主要的收入來源來源之一，層出不窮的網(wǎng)絡廣告投放技術導致了廣告投放的無目的性，投放的廣告很大程度上是用戶不感興趣的，這不僅不會引起用戶的興趣而去點擊廣告，還會招致用戶的厭煩。針對這種情況，出現(xiàn)了精準廣告投放這一概念。網(wǎng)絡廣告發(fā)展的精準化趨勢催生了行為定向網(wǎng)絡廣告技術，這種廣告模式具有精準性、及時性、高效性，從而備受人們的關注。行為定向廣告的原理是挖掘用戶瀏覽網(wǎng)頁記錄中包含的用戶行為特征來投放廣告，根據(jù)用戶的行為特征，，廣告主可以更精確地對目標用戶進行廣告投放,用最少的投資換取最高的利潤。由于定向廣告具有覆蓋面廣、精準度高的特點,被各大互聯(lián)網(wǎng)公司和精準化廣告服務商所重視。 本文首先介紹了行為定向網(wǎng)絡廣告研究的背景和意義，行為定向廣告的特點和投放過程中的難點，引出了本文的研究內(nèi)容和創(chuàng)新點，之后介紹了廣告投放精準化技術的現(xiàn)狀。論述了網(wǎng)頁預處理的過程，文檔聚類常用的方法和常用的文檔統(tǒng)計模型。然后針對行為定向廣告，提出了一種基于LDA的行為定向廣告算法的研究。該算法首先根據(jù)用戶的歷史訪問記錄，引入了追蹤窗口的概念,根據(jù)網(wǎng)頁訪問時間的先后順序進行排列,建立時變用戶行為模型利用改進的k-means算法來計算用戶的長期行為和短期行為；然后由時變用戶行為模型結(jié)合LDA主題模型，提出了一種高效的行為定向廣告投放算法LDA_M。最后根據(jù)LDA_M算法得到的最能代表用戶行為的幾個關鍵詞與廣告關鍵詞進行匹配，選出和用戶行為特征最相關的幾個廣告進行投放。該算法解決了傳統(tǒng)廣告投放算法不能分析語義和LDA模型廣告投放算法不能很好的挖掘用戶行為變化的缺點。實驗表明，該算法能很好的區(qū)分用戶的長期行為和短期行為，精確地挖掘用戶的行為特征，投放與用戶行為特征相關的廣告，提高了廣告投放的精確度。
[Abstract]:With the development of electronic commerce, the network media advertisement has become one of the main revenue sources of the website. The endless network advertising technology has led to the aimlessness of the advertisement, and the advertisement placed is to a large extent uninteresting to the users. This will not only not arouse the interest of users to click on advertising, but will cause users to be bored. In response to this situation, the concept of precision advertising has emerged. The trend of precision in the development of online advertising has given birth to behavioral orientation online advertising technology, this advertising model has precision, timeliness and efficiency, so people pay close attention to it. The principle of behavior oriented advertising is to excavate the user behavior characteristics contained in the user browsing web pages to place advertisements. According to the behavior characteristics of users, advertisers can more precisely place advertisements on the target users. The least investment in exchange for the highest profit. Because of its wide coverage and high accuracy, targeted advertising has been paid attention to by major Internet companies and precision advertising service providers. This paper first introduces the background and significance of the research on behavioral orientation online advertising, the characteristics of behavioral targeted advertising and the difficulties in the delivery process, leads to the research content and innovation of this paper, and then introduces the present situation of the precision advertising technology. This paper discusses the process of web page preprocessing, the common methods of document clustering and the commonly used document statistical model. Then, a behavioral orientation advertising algorithm based on LDA is proposed. The algorithm firstly introduces the concept of tracking window according to the user's historical access record, and arranges it according to the order of page access time. A time-varying user behavior model is established to calculate the long-term and short-term behaviors of users using an improved k-means algorithm, and then an efficient behavioral targeting advertising algorithm LDAS is proposed by combining the time-varying user behavior model with the LDA topic model. Finally, according to the LDA_M algorithm, the most representative of the user behavior of several keywords and advertising keyword matching, select the most relevant to the user behavior characteristics of several ads to put in. This algorithm solves the shortcomings that the traditional advertising algorithm can not analyze the semantics and the LDA model advertising algorithm can not well mining user behavior changes. Experiments show that the algorithm can distinguish the long-term behavior and short-term behavior of users accurately mining the behavior characteristics of users and putting in advertisements related to user behavior characteristics and improves the accuracy of advertising.
【學位授予單位】：遼寧大學
【學位級別】：碩士
【學位授予年份】：2014
【分類號】：TP393.092

【參考文獻】

相關期刊論文 前1條

1 顧益軍,樊孝忠,王建華,汪濤,黃維金;中文停用詞表的自動選取[J];北京理工大學學報;2005年04期

相關博士學位論文 前1條

1 陳冬玲;基于潛在語義的個性化搜索關鍵技術研究[D];東北大學;2009年

本文編號：1847398