Adaptive targeting for finding look-alike users

Patent metadata

Field	Value
Publication number	US-9087332-B2
Application number	US-87177510-A
Country	US
Kind code	B2
Filing date	Aug 30, 2010
Priority date	Aug 30, 2010
Publication date	Jul 21, 2015
Grant date	Jul 21, 2015

Abstract

A method for adaptive display of internet advertisements to look-alike users using a desired user profile dataset as a seed to machine learning modules. Upon availability of a desired user profile, that user profile is mapped other look-alike users (from a larger database of users). The method proceeds to normalize the desired user profile object, proceeds to normalize known user profile objects, then seeding a machine-learning training model with the normalized desired user profile object. A scoring engine uses the normalized user profiles for matching based on extracted features (i.e. extracted from the normalized user profile objects). Once look-alike users have been identified, the internet display system may serve advertisements to the look-alike users, and analyze look-alike users' behaviors for storing the predicted similar user profile objects into the desired user profile object dataset, thus adapting to changing user behavior.

First claim

We claim:

1. A method for adaptive display of an advertisement to look-alike users using a desired user profile dataset, the method comprising:
obtaining, by a computer, a plurality of known user profiles of known users who have been recorded to interact with an advertiser, wherein each of the plurality of known user profiles includes:
historical components reflecting a stream of events of the known user prior to a current time, and
a temporary component reflecting a state of the known user at the current time;

automatically creating, by a computer, a plurality of desired user profiles of desired users who are not included in the plurality of known user profiles, wherein each of the plurality of the desired user profiles includes:
historical components reflecting a stream of events of the desired user prior to the current time, and
a temporary component reflecting a state of the desired user at the current time;

scoring, by a computer with a machine-learned model, similarities between the plurality of desired user profiles with the plurality of known user profiles based on the temporal component of the plurality of known user profile and the temporal component of the plurality of desired user profile for adapting to changes of user behavior;
selecting, by a computer, a predicted user from the desired users based on the score of the plurality of desired user profile and; and
serving, by a computer, an advertisement to the predicted user.

2. The method of claim 1 , wherein the similarity between a desired user and a known user comprises a similarity between a current component of the desired user and a historical component of the known user.

3. The method of claim 1 , wherein the similarity between a desired user and a known user comprises a similarity between a current component of the desired user and a current component of the known user.

4. The method of claim 1 , wherein the similarity between a desired user and a known user comprises a similarity between a historical component of the desired user and a historical component of the known user.

5. An advertising server, comprising at least one processor and processor-readable storage medium, wherein the storage medium comprise a set of instructions for adaptive display of an advertisement to look-alike users using a desired user profile dataset, and wherein when executing the set of instructions, the processor is directed to:
obtain a plurality of known user profiles of known users who have been recorded to interact with an advertiser, wherein each of the plurality of known user profiles includes:
historical components reflecting a stream of events of the known user prior to a current time, and
a temporary component reflecting a state of the known user at the current time;

automatically create a plurality of desired user profiles of desired users who are not included in the plurality of known user profiles, wherein each of the plurality of the desired user profiles includes:
historical components reflecting a stream of events of the desired user prior to the current time, and
a temporary component reflecting a state of the desired user at the current time;

score through a machine-learned model similarities between the plurality of desired user profiles with the plurality of known user profiles based on the temporal component of the plurality of known user profile and the temporal component of the plurality of desired user profile for adapting to changes of user behavior;
select, by a computer, a predicted user from the desired users based on the score of the plurality of desired user profile and; and
serve an advertisement to the predicted user.

6. The advertising server of claim 5 , wherein the similarity between a desired user and a known user comprises a similarity between a current component of the desired user and a current component of the known user.

7. The advertising server of claim 5 , wherein the similarity between a desired user and a known user comprises a similarity between a current component of the desired user and a historical component of the known user.

8. The advertising server of claim 5 , wherein the processor is directed to score the similarities based on a linear model.

9. The advertising server of claim 5 , wherein the processor is directed to score the similarities based on a clustering model.

10. The advertising server of claim 5 , wherein the processor is directed to score the similarities based on a classifier.

11. A non-transitory computer readable medium comprising a set of instructions for adaptive display of an advertisement to look-alike users using a desired user profile dataset which, when executed by a computer, cause the computer to perform actions of:
obtaining a plurality of known user profiles of known users who have been recorded to interact with an advertiser, wherein each of the plurality of known user profiles includes:
historical components reflecting a stream of events of the known user prior to a current time, and
a temporary component reflecting a state of the known user at the current time;

automatically creating a plurality of desired user profiles of desired users who are not included in the plurality of known user profiles, wherein each of the plurality of the desired user profiles includes
historical components reflecting a stream of events of the desired user prior to the current time, and
a temporary component reflecting a state of the desired user at the current time;

scoring, with a machine-learned model, similarities between the plurality of desired user profiles with the plurality of known user profiles based on the temporal component of the plurality of known user profile and the temporal component of the plurality of desired user profile for adapting to changes of user behavior;
selecting, by a computer, a plurality of predicted users from the desired users based on the score of the plurality of desired user profile and; and
serving an advertisement to the predicted user.

12. The non-transitory computer readable medium of claim 11 , wherein the similarity between a desired user and a known user comprises a similarity between a current component of the desired user and a current component of the known user.

13. The non-transitory computer readable medium of claim 11 , wherein the similarity between a desired user and a known user comprises a similarity between a current component of the desired user and a historical component of the known user.

14. The non-transitory computer readable medium of claim 11 , wherein the scoring of the similarities is based on a linear model.

15. The non-transitory computer readable medium of claim 11 , wherein the scoring of the similarities is based on a clustering model.

16. The non-transitory computer readable medium of claim 11 , wherein the scoring of the similarities is based on a classifier.

Assignees

• Bagherjeiran Abraham
• Tang Renjie
• Zhang Zengvan
• Hatch Andrew
• Ratnaparkhi Adwait
• Parekh Ralesh
• Yahoo Inc

Inventors

• Bagherjeiran Abraham
• Tang Renjie
• Zhang Zengvan
• Hatch Andrew
• Ratnaparkhi Adwait
• Parekh Ralesh

Classifications

• G06Q30/0255

  based on user history · CPC title

• G06Q30/0269

  based on user profile or attribute · CPC title

• G06Q30/00 Primary

  Commerce · CPC title

• G06Q30/0251

  Targeted advertisements · CPC title