Behavioral targeting system

• US 8,504,575 B2
• Filed: 03/29/2006
• Issued: 08/06/2013
• Est. Priority Date: 03/29/2006
• Status: Active Grant

• Current Assignee: Twitter Inc (X Holdings Corp.)
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Koran, Joshua M., Chung, Christina Yip, Gupta, Abhinav, John, George H., Yin, Hongfeng, Lin, Long-Ji, Frankel, Richard
• Primary Examiner(s): Hoang, Son T
• Application Number:
  US11/394,343
• Publication Number:
  US 20070260596A1
• Time in Patent Office: 2,687 Days
• Field of Search: 705/748, 705/758
• US Class Current: 707/748
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/337 Profile generation, learning or modification
• CPC Class Codes: G06F 16/635 Filtering based on additional data, e.g. user or group profiles
• CPC Class Codes: G06F 16/84 Mapping; Conversion
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06Q 10/067 Enterprise or organisation modelling
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0255 based on user history

Abstract

A behavioral targeting system determines user profiles from online activity. The system includes a plurality of models that define parameters for determining a user profile score. Event information, which comprises on-line activity of the user, is received at an entity. To generate a user profile score, a model is selected. The model comprises recency, intensity and frequency dimension parameters. The behavioral targeting system generates a user profile score for a target objective, such as brand advertising or direct response advertising. The parameters from the model are applied to generate the user profile score in a category. The behavioral targeting system has application for use in ad serving to on-line users.

Claims

1. A method, implemented by at least one computer processor, for determining a user profile from online activity, the method comprising:

processing a user data set, comprising event information from a plurality of events, compiled from past on-line activity between users of the user data set and an entity;

analyzing the user data set to ascertain a level of performance of the event information to predict user interest in each of a plurality of categories, wherein a category specifies a subject matter;

generating a plurality of models, one for each of the plurality of categories, wherein each model comprises a plurality of weights for determining a user interest score in a corresponding category;

generating weights for the plurality of models by ascribing a predictive value to a plurality of types of the event information in accordance with the level of performance of a particular type of the event information to predict the user interest in a corresponding category;

storing the plurality of models at the entity for the plurality of categories;

receiving, at the entity, the event information from at least one event from a user;

classifying the event information in a particular category of the plurality of categories;

identifying a type of the received event information;

selecting a model, based on the particular category, to generate at least a user profile score for the particular category; and

generating the at least one user profile score for the particular category by applying at least one weight based on the type of the received event information and the particular category from the model selected, wherein the user profile score indicates the user interest in the subject matter of the particular category.

2. The method as set forth in claim 1, further comprising:

adding at least one new model comprising new parameters and rules so as to provide extensibility;

receiving, at the entity, the event information;

selecting the new model to generate the user profile score; and

generating the at least one user profile score in the particular category by applying the new parameters and rules from the new model selected to the user event information.

3. The method as set forth in claim 1, further comprising serving an advertisement to the user based on the user interest score.

4. The method as set forth in claim 1, wherein the plurality of weights comprise recency dimension weights that provide input to the user interest score based on how recent the event information occurred.

5. The method as set forth in claim 1, wherein the plurality of weights comprise intensity dimension weights that provides input to the user interest score based on effectiveness of the event information to predict the user interest in the particular category.

6. The method as set forth in claim 1, wherein the plurality of weights comprise frequency dimension weights that provides input to the user interest score based on frequency of occurrence of the event information.

7. The method as set forth in claim 1, wherein:

selecting the model to generate the user score comprises selecting a model based on a targeting objective; and

generating the at least one user interest score comprises generating a user interest score for the targeting objective.

8. The method as set forth in claim 7, wherein the targeting objective comprises brand advertising.

9. The method as set forth in claim 7, wherein the targeting objective comprises direct response advertising.

10. A system for determining user interest from online activity, the system comprising:

at least one server computer is configured for;

processing a user data set, comprising event information from a plurality of events, compiled from past on-line activity between users of the user data set and an entity,

analyzing the user data set to ascertain a level of performance of the event information to predict user interest in each of a plurality of categories, wherein a category specifies a subject matter,

generating a plurality of models, one for each of the plurality of categories, wherein each model comprises a plurality of weights for determining a user interest score in a corresponding category, and

generating weights for the plurality of models by ascribing a predictive value to a plurality of types of the event information in accordance with the level of performance of a particular type of the event information to predict the user interest in a corresponding category;

at least one storage device for storing the plurality of models at the entity for the plurality of categories;

the server computer is further configured for;

receiving, at the entity, the event information from at least one event from a user,

classifying the event information in a particular category of the plurality of categories,

identifying a type of the received event information, for selecting a model, based on the particular category, to generate at least a user profile score for the particular category, and

generating the at least one user profile score for the particular category by applying at least one weight based on the type of the received event information and the particular category from the model selected, wherein the user profile score indicates the user interest in the subject matter of the particular category.

11. The system as set forth in claim 10, wherein the system comprises an extensible system, such that at least one additional model is added without requiring change to the system.

12. The system as set forth in claim 10, further comprising an advertisement server computer for serving an advertisement to the user based on the user interest score.

13. The system as set forth in claim 10, wherein the plurality of weights comprise recency dimension weights that provide input to the user interest score based on how recent the event information occurred.

14. The system as set forth in claim 10, wherein the plurality of weights comprise intensity dimension weights that provides input to the user interest score based on effectiveness of the event information to predict user interest in the particular category.

15. The system as set forth in claim 10, wherein the plurality of weights comprise frequency dimension weights that provides input to the user interest score based on frequency of occurrence of the event information.

16. The system as set forth in claim 10, wherein the server computer is further configured for generating the user profile score for a target objective.

17. The system as set forth in claim 16, wherein the target objective comprises brand advertising.

18. The system as set forth in claim 16, wherein the target objective comprises direct response advertising.

19. A non-transitory computer readable storage medium comprising a set of instructions which, when executed by a computer, causes the computer to determine a user profile from online activity, the set of instructions for:

processing a user data set, comprising event information from a plurality of events, compiled from past on-line activity between users of the user data set and an entity;

analyzing the user data set to ascertain a level of performance of the event information to predict user interest in each of a plurality of categories, wherein a category specifies a subject matter;

generating a plurality of models, one for each of the plurality of categories, wherein each model comprises a plurality of weights for determining a user interest score in a corresponding category;

generating weights for the plurality of models by ascribing a predictive value to a plurality of types of the event information in accordance with the level of performance of a particular type of the event information to predict the user interest in a corresponding category;

storing the plurality of models at the entity for the plurality of categories;

receiving, at the entity, the event information from at least one event from a user;

classifying the event information in a particular category of the plurality of categories;

identifying a type of the received event information;

selecting a model, based on the particular category, to generate at least a user profile score for the particular category; and

generating the at least one user profile score for the particular category by applying at least one weight based on the type of the received event information and the particular category from the model selected, wherein the user profile score indicates the user interest in the subject matter of the particular category.

20. The non-transitory computer readable storage medium as set forth in claim 19, further comprising serving an advertisement to the user based on the user profile score.

21. The non-transitory computer readable storage medium as set forth in claim 19, wherein the plurality of weights comprise recency dimension weights that provide input to the user interest score based on how recent the event information occurred.

22. The non-transitory computer readable storage medium as set forth in claim 19, wherein the plurality of weights comprise intensity dimension weights that provides input to the user interest score based on effectiveness of the event information to predict the user interest in the particular category.

23. The non-transitory computer readable storage medium as set forth in claim 19, wherein the plurality of weights comprise frequency dimension weights that provides input to the user interest score based on frequency of occurrence of the event information.

24. The non-transitory computer readable storage medium as set forth in claim 19, wherein:

selecting the model to generate the user score comprises selecting a model based on a target objective; and

generating the at least one user profile score comprises generating a user profile score for the target objective.