在線關(guān)鍵字拍賣Agent競價策略研究
發(fā)布時間：2018-09-04 09:20
【摘要】：隨著互聯(lián)網(wǎng)的發(fā)展，，搜索引擎已經(jīng)成為人們快速搜索信息的重要工具，而關(guān)鍵字廣告作為搜索引擎的重要經(jīng)濟基礎(chǔ)之一，有效地滿足了廣告主的營銷需要，同時也給搜索引擎提供商帶來巨大的利潤。關(guān)鍵字廣告拍賣不僅成為微觀經(jīng)濟理論在互聯(lián)網(wǎng)上最成功的應(yīng)用之一，而且推動了科學(xué)領(lǐng)域的眾多學(xué)者對其背后機制的研究，成為電子商務(wù)領(lǐng)域的研究熱點。交易智能體競賽，簡稱TAC，是由卡耐基梅隆大學(xué)等學(xué)校聯(lián)合主辦，旨在模擬真實的市場行為，將目前人工智能的理論研究成果應(yīng)用到現(xiàn)實的交易過程中。TAC/AA是虛擬關(guān)鍵字廣告拍賣平臺，研究者可以將研究成果應(yīng)用到該平臺上，驗證算法的有效性。本文圍繞TAC/AA平臺設(shè)計了一種能夠制定連續(xù)最優(yōu)競價策略的Agent模型，并提出了動態(tài)多樣精英PSO算法，有效解決了Agent模型中優(yōu)化問題的局部收斂問題。 首先本文分析了國內(nèi)外關(guān)鍵字拍賣的研究現(xiàn)狀，介紹了TAC平臺的設(shè)計思想和比賽的規(guī)則，然后研究了關(guān)鍵字拍賣理論、Agent理論和進化算法的相關(guān)理論。針對TAC平臺的關(guān)鍵字拍賣競賽，設(shè)計了TAC-HEU-AA(THA)Agent模型，分析了預(yù)測器的算法的效率，通過實驗討論了各個模塊的必要性，并將THA模型與參與TAC/AA決賽的Agent進行比較試驗，驗證了模型的有效性，分析了模型的優(yōu)缺點。針對THA模型中的優(yōu)化器的多選擇背包問題(MCKP，Multi-Choice Knapsack Problem)，提出了一種新的基于多樣精英選擇策略的粒子群算法(DME-PSO)。定義了3種粒子的運動趨勢和4種粒子間的運動行為。根據(jù)粒子的運動行為選擇粒子以保持種群的多樣性。在變異過程中，描述了種群的精英飽和現(xiàn)象，在種群精英飽和狀態(tài)時對種群加入擾動。在實驗中，將DME-PSO算法與貪婪算法，多種群遺傳算法和加入高斯擾動的粒子群算法進行比較。實驗表明DME-PSO算法在物品數(shù)量增加時表現(xiàn)出較好的優(yōu)化效果，從而更好的解決組合優(yōu)化的局部收斂問題。
[Abstract]:With the development of Internet, search engine has become an important tool for people to search for information quickly. As one of the important economic foundations of search engine, keyword advertisement meets the marketing needs of advertisers effectively. At the same time, it also brings huge profits to search engine providers. Keyword advertising auction has not only become one of the most successful applications of microeconomic theory on the Internet, but also promoted the research of the mechanism behind it by many scholars in the field of science, and has become a research hotspot in the field of electronic commerce. The transaction Agent Competition, or TAC, is sponsored by schools such as Carnegie Mellon University to simulate real market behavior. Applying the current theoretical research results of artificial intelligence to real transactions. TAC / AA is a virtual keyword advertising auction platform. Researchers can apply the research results to this platform to verify the effectiveness of the algorithm. This paper designs a Agent model based on TAC/AA platform, which can formulate continuous optimal bidding strategy, and proposes dynamic and diverse elite PSO algorithm, which effectively solves the problem of local convergence of optimization problem in Agent model. Firstly, this paper analyzes the research status of keyword auction at home and abroad, introduces the design idea of TAC platform and the rules of competition, and then studies the theory of keyword auction and the related theories of evolutionary algorithm. Aiming at the keyword auction competition of TAC platform, this paper designs the TAC-HEU-AA (THA) Agent model, analyzes the efficiency of the algorithm of the predictor, discusses the necessity of each module through experiments, and compares the THA model with the Agent participating in the TAC/AA finals. The validity of the model is verified and the advantages and disadvantages of the model are analyzed. Aiming at the multi-selection knapsack problem of optimizer in THA model (MCKP,Multi-Choice Knapsack Problem), a new particle swarm optimization algorithm (DME-PSO) based on multiple elitist selection strategies is proposed. The movement trends of three kinds of particles and the motion behavior of four kinds of particles are defined. The particles are selected according to their motion behavior to maintain the diversity of the population. In the process of variation, the phenomenon of elite saturation of population is described, and the disturbance is added to the population in the state of elite saturation. In the experiment, DME-PSO algorithm is compared with greedy algorithm, multi-population genetic algorithm and particle swarm optimization algorithm with Gao Si disturbance. The experimental results show that the DME-PSO algorithm performs well when the number of items increases, thus solving the problem of local convergence of combinatorial optimization better.
【學(xué)位授予單位】：哈爾濱工程大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2012
【分類號】：TP391.3;TP18
本文編號：2221659
[Abstract]:With the development of Internet, search engine has become an important tool for people to search for information quickly. As one of the important economic foundations of search engine, keyword advertisement meets the marketing needs of advertisers effectively. At the same time, it also brings huge profits to search engine providers. Keyword advertising auction has not only become one of the most successful applications of microeconomic theory on the Internet, but also promoted the research of the mechanism behind it by many scholars in the field of science, and has become a research hotspot in the field of electronic commerce. The transaction Agent Competition, or TAC, is sponsored by schools such as Carnegie Mellon University to simulate real market behavior. Applying the current theoretical research results of artificial intelligence to real transactions. TAC / AA is a virtual keyword advertising auction platform. Researchers can apply the research results to this platform to verify the effectiveness of the algorithm. This paper designs a Agent model based on TAC/AA platform, which can formulate continuous optimal bidding strategy, and proposes dynamic and diverse elite PSO algorithm, which effectively solves the problem of local convergence of optimization problem in Agent model. Firstly, this paper analyzes the research status of keyword auction at home and abroad, introduces the design idea of TAC platform and the rules of competition, and then studies the theory of keyword auction and the related theories of evolutionary algorithm. Aiming at the keyword auction competition of TAC platform, this paper designs the TAC-HEU-AA (THA) Agent model, analyzes the efficiency of the algorithm of the predictor, discusses the necessity of each module through experiments, and compares the THA model with the Agent participating in the TAC/AA finals. The validity of the model is verified and the advantages and disadvantages of the model are analyzed. Aiming at the multi-selection knapsack problem of optimizer in THA model (MCKP,Multi-Choice Knapsack Problem), a new particle swarm optimization algorithm (DME-PSO) based on multiple elitist selection strategies is proposed. The movement trends of three kinds of particles and the motion behavior of four kinds of particles are defined. The particles are selected according to their motion behavior to maintain the diversity of the population. In the process of variation, the phenomenon of elite saturation of population is described, and the disturbance is added to the population in the state of elite saturation. In the experiment, DME-PSO algorithm is compared with greedy algorithm, multi-population genetic algorithm and particle swarm optimization algorithm with Gao Si disturbance. The experimental results show that the DME-PSO algorithm performs well when the number of items increases, thus solving the problem of local convergence of combinatorial optimization better.
【學(xué)位授予單位】：哈爾濱工程大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2012
【分類號】：TP391.3;TP18
【參考文獻】
相關(guān)期刊論文 前5條
1 鮑江宏;楊啟貴;;多選擇背包問題的快速求解算法[J];華南理工大學(xué)學(xué)報(自然科學(xué)版);2009年04期
2 戎文晉;劉樹林;;關(guān)鍵詞拍賣中最優(yōu)保留價的研究[J];管理科學(xué)學(xué)報;2010年04期
3 王勇;蔡自興;周育人;肖赤心;;約束優(yōu)化進化算法[J];軟件學(xué)報;2009年01期
4 汪定偉;;考慮廣告商信譽的搜索引擎排位拍賣的機制設(shè)計[J];系統(tǒng)工程理論與實踐;2011年01期
5 艾熱提·阿布力克木;劉芳園;;基于主體Agent上的意圖表示[J];新疆師范大學(xué)學(xué)報(自然科學(xué)版);2009年04期
相關(guān)博士學(xué)位論文 前1條
1 宋海剛;開放電子市場中的議價協(xié)議及議價策略研究[D];華中科技大學(xué);2005年
本文編號：2221659
本文鏈接：http://sikaile.net/kejilunwen/sousuoyinqinglunwen/2221659.html
最近更新
教材專著
