Presentation is loading. Please wait.

Presentation is loading. Please wait.

Personalized Delivery of On-Line Search Advertisement Based on User Interests Guangyi Xiao, Zhiguo Gong University of Macau.

Similar presentations


Presentation on theme: "Personalized Delivery of On-Line Search Advertisement Based on User Interests Guangyi Xiao, Zhiguo Gong University of Macau."— Presentation transcript:

1 Personalized Delivery of On-Line Search Advertisement Based on User Interests Guangyi Xiao, Zhiguo Gong University of Macau

2 Advertising in Google II

3 Search advertising becomes the main revenue share  Internet advertising revenues (“revenues”) in the United States $23 billion in 2008, 10.6% increase comparing with 2007  Despite a difficult U.S. economy, internet advertising’s continued growth  Search 45% in 2008, 42% in 2007

4 Object of advertising auction  maximize revenues,  efficiency  and user satisfaction.

5 Payment scheme in online advertising  cost-per-thousands-impressions (CPM) model  cost-per- click (CPC) model  cost-per-event (CPE) model (e.g. Sales)

6 Two kinds of ranked framework for advertising  Ranked by bid, adopted previously by Yahoo  Ranked by revenue, adopted by Google.  In this paper, maximizing revenue is the objective for advertisement scheduling.

7 Three Auction Scheme  First Price Scheme (It is very simple.)  Second Price Scheme, used in our system.  VCG (Vickery Clarke Groves), charge bidder i according to the overall loss of all others because of i’s presence

8 Advertisement Ranking  Ranked by bid value  A1>A2>…>An  Ignore if the adv be clicked  Ranked by bid value, together with ads quality  A 1 Q 1 >A 2 Q 2 >…>A n Q n  Ignore if the individual user clicks it  Ranked by bid value, together with user’s interest  A 1 I 1 >A 2 I 2 >…>A n I n

9 Revenue  Limited space for advertising  Select only several ads to display  Second price scheme is used  Only m ads can be displayed

10 Revenue (cont.)

11 Interest Model 1.f 0 (u,j), user u’s query sessions comparing ads j’s bid terms 2.f 1 (u,j), user u’s overall history query terms comparing with j’s bid terms 3.f 2 (u,j), user u’s view of j in the history 4.f 3 (u,j), user u’s click of j in the history

12 Experiment

13 An Example of Our Ads Scheduling


Download ppt "Personalized Delivery of On-Line Search Advertisement Based on User Interests Guangyi Xiao, Zhiguo Gong University of Macau."

Similar presentations


Ads by Google