Visitor profile modeling

• US 8,190,475 B1
• Filed: 09/04/2008
• Issued: 05/29/2012
• Est. Priority Date: 09/05/2007
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Merrill, John W.
• Primary Examiner(s): Weiss, John
• Assistant Examiner(s): Chornesky, Adam
• Application Number:
  US12/204,679
• Time in Patent Office: 1,363 Days
• Field of Search: 705/14.66, 705/14.53, 705/14.49, 705/10, 705/14.52, 707/5, 707/E17.084, 707/6, 709/217
• US Class Current: 705/14.49
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data
• CPC Class Codes: G06Q 30/0251 Targeted advertisements
• CPC Class Codes: G06Q 30/0254 based on statistics
• CPC Class Codes: G06Q 30/0255 based on user history
• CPC Class Codes: Y10S 707/99936 Pattern matching access

Abstract

Methods, systems and computer program products for inferring a value of a visitor property for a visitor to a web destination are described. A set of traversals indicating visits to web destinations are received, including scored and unscored traversals. Initial destination scores for the web destinations visited by the scored traversals are inferred based on the known scores of the scored traversals. Initial scores for the unscored traversals are estimated based on the initial destinations scores. Revised destination scores for the web destinations visited by the scored traversals and initial destination scores for the web destinations only visited by the unscored traversals are inferred. Revised scores for the unscored traversals are estimated based on the revised destination scores and initial destination scores. Revised destination scores for all of the web destinations are inferred based on the scores of the scored traversals and the revised scores of the unscored traversals.

Claims

1. A computer implemented method comprising:

receiving a set of traversals indicating visits to a set of web destinations, the set of traversals including scored and unscored traversals where a scored traversal is associated with a known score of at least one visitor property of a visitor associated with the traversal;

inferring initial destination scores of the at least one visitor property for the web destinations visited by the scored traversals based on the known scores of the scored traversals;

estimating initial scores for the unscored traversals;

inferring revised destination scores for the web destinations visited by the scored traversals and inferring initial destination scores for the web destinations only visited by the unscored traversals based on the known scores of the scored traversals and the initial scores of the unscored traversals;

estimating, by one or more processors, revised scores for the unscored traversals based on the revised destination scores for the web destinations visited by the scored traversals and the initial destination scores for the web destinations only visited by the unscored traversals;

inferring, by the one or more processors, revised destination scores for all of the web destinations based on the known scores of the scored traversals and the revised scores of the unscored traversals; and

performing a plurality of iterations of the estimating and inferring steps above.

2. The method of claim 1, wherein the plurality of iterations are performed until the estimated revised scores of the unscored traversals stabilize.

3. The method of claim 2, wherein a revised score stabilizes if the root mean square difference between the score for two consecutive iterations is less than a predetermined threshold value.

4. The method of claim 1, wherein the at least one visitor property comprises a visitor demographic property.

5. A computer implemented method comprising:

receiving a set of traversals indicating visits to a set of web destinations, the set of traversals including scored and unscored traversals where a scored traversal is associated with a score of at least one visitor property of a visitor associated with the traversal;

determining destination scores for the web destinations based on the scores for the set of traversals;

after determining the destination scores, determining initial scores for the unscored traversals;

resealing, by one or more processors, the determined destination scores based at least in part on the initial scores for the unscored traversals;

determining, by the one or more processors, revised traversal scores for the unscored traversals based on the resealed destination scores; and

resealing, by the one or more processors, the revised traversal scores based at least in part on the revised traversal scores for the unscored traversals.

6. The method of claim 5, further comprising:

determining if the revised traversal scores have stabilized; and

when the revised traversal scores have not stabilized, then repeating the determining destination scores steps through the resealing the revised traversal scores step until the revised traversal scores have stabilized.

7. The method of claim 5, wherein resealing the determined destination scores comprises:

linearly resealing the determined destination scores to adjust a mean and variance of the determined destination scores when computed from the unscored traversals to a mean and variance of the destination scores when computed from the scored traversals.

8. The method of claim 5, wherein resealing the revised traversal scores comprises:

linearly resealing the revised traversal scores to adjust the mean of the revised traversal scores attributed to the unscored traversals to that of the traversal scores attributed to the scored traversals, while increasing the variance of the revised scores attributed to the unscored traversals.

9. The method of claim 5, wherein the determining the destination score steps through the determining revised traversal scores step are performed by an algorithm that infers the destination and traversal scores by imputing values for a set of common latent variables.

10. A system comprising:

a processor;

a storage device coupled to the processor and configurable for storing instructions, which, when executed by the processor, cause the processor to perform operations comprising;

receiving a set of traversals indicating visits to a set of web destinations, the set of traversals including scored and unscored traversals where a scored traversal is associated with a known score of at least one visitor property of a visitor associated with the traversal;

inferring initial destination scores of the at least one visitor property for the web destinations visited by the scored traversals based on the known scores of the scored traversals;

estimating initial scores for the unscored traversals;

inferring revised destination scores for the web destinations visited by the scored traversals and inferring initial destination scores for the web destinations only visited by the unscored traversals based on the known scores of the scored traversals and the initial scores of the unscored traversals;

estimating revised scores for the unscored traversals based on the revised destination scores for the web destinations visited by the scored traversals and the initial destination scores for the web destinations only visited by the unscored traversals;

inferring revised destination scores for all of the web destinations based on the known scores of the scored traversals and the revised scores of the unscored traversals; and

performing a plurality of iterations of the estimating and inferring steps above.

11. The system of claim 10, wherein the plurality of iterations are performed until the estimated revised scores of the unscored traversals stabilize.

12. The system of claim 11, wherein a revised score stabilizes if the root mean square difference between the score for two consecutive iterations is less than a predetermined threshold value.

13. The system of claim 10, wherein the at least one visitor property comprises a visitor demographic property.

14. A computer readable medium having instructions encoded thereon, which, when executed by a processor, cause the processor to perform operations comprising:

receiving a set of traversals indicating visits to a set of web destinations, the set of traversals including scored and unscored traversals where a scored traversal is associated with a known score of at least one visitor property of a visitor associated with the traversal;

inferring initial destination scores of the at least one visitor property for the web destinations visited by the scored traversals based on the known scores of the scored traversals;

estimating initial scores for the unscored traversals;

inferring revised destination scores for the web destinations visited by the scored traversals and inferring initial destination scores for the web destinations only visited by the unscored traversals based on the known scores of the scored traversals and the initial scores of the unscored traversals;

estimating revised scores for the unscored traversals based on the revised destination scores for the web destinations visited by the scored traversals and the initial destination scores for the web destinations only visited by the unscored traversals;

inferring revised destination scores for all of the web destinations based on the known scores of the scored traversals and the revised scores of the unscored traversals; and

performing a plurality of iterations of the estimating revised scores and inferring revised destination scores steps above.

15. The computer readable medium of claim 14, wherein the plurality of iterations are performed until the estimated revised scores of the unscored traversals stabilize.

16. The computer readable medium of claim 15, wherein a revised score stabilizes if the root mean square difference between the score for two consecutive iterations is less than a predetermined threshold value.

17. The computer readable medium of claim 14, wherein the at least one visitor property comprises a visitor demographic property.