Systems and methods for generating expanded user segments

• US 8,655,695 B1
• Filed: 05/07/2010
• Issued: 02/18/2014
• Est. Priority Date: 05/07/2010
• Status: Active Grant

• Current Assignee: Yahoo AD Tech LLC
• Original Assignee: AOL Advertising Incorporated (Apollo Global Management, Inc.)
• Inventors: Qu, Yan, Wang, Jing, Sun, Yang, Holtan, Hans Marius
• Primary Examiner(s): Nguyen, Nga B.
• Application Number:
  US12/775,826
• Time in Patent Office: 1,383 Days
• Field of Search: 705/7, 705/10, 705/14
• US Class Current: 705/7
• CPC Class Codes: G06Q 30/0251 Targeted advertisements

Abstract

Systems and methods are disclosed for generating an expanded user segment based on a target segment of users associated with a specified target feature. In one implementation, a method is provided that includes receiving information about online activity by a set of users, the information including a unique user ID associated with each of the set of users; extracting from the received information, data about the set of users who are associated with the specified target feature; specifying one or more independent features relevant to a user association with the specified target feature; building a model that represents the probability of any user being associated with the specified target feature, based on the extracted data and the independent features; and using the model to score a network population of users relative to the specified feature, based on information received about online activity by the network population of users.

Claims

1. A computer-implemented method for generating an expanded user segment based on a target segment of users, the method comprising the following steps performed by one or more processors:

receiving information about online activity by a first plurality of users, the information including a unique user ID associated with each of the first plurality of users;

extracting, from the received information, data about the first plurality of users who are associated with a plurality of specified target features;

selecting a combination of two or more of the plurality of specified target features as an output for building at least one model;

building the model that represents the probability of any user being associated with the selected combination of the plurality of specified target features, based on the extracted data about the first plurality of users;

using the model to score a second plurality of users relative to the selected combination of the plurality of specified target features, based on information received about online activity by the second plurality of users; and

generating an expanded user segment by grouping users who have been scored above a threshold value for the selected combination of the plurality of specified target features.

2. The computer-implemented method of claim 1, wherein the information about online activity is obtained through one or more cookies.

3. The computer-implemented method of claim 1, wherein the received information includes one or more of:

user profiles, cookie data, behavioral logs, leadback logs, click logs, action logs, and impression logs.

4. The computer-implemented method of claim 1, wherein the plurality of specified target features includes one or more of:

whether a user provided certain demographic information, whether a user clicked on a particular advertisement, and whether a user is likely to convert to a particular campaign.

5. The computer-implemented method of claim 1, wherein the plurality of specified target features includes one or more of:

a demographic value, a geographic location, a response to a campaign, a conversion, a view of a web page, and a click of an ad.

6. The computer-implemented method of claim 1, wherein the step of building a model is formulated as a supervised learning task.

7. The computer-implemented method of claim 1, wherein the model is built based on one or more of:

random forest modeling, linear regression, logistic regression for continuous output values, and Naï

ve Bayesian classification.

8. The computer-implemented method of claim 1, wherein the model is built based on a list of independent features identified as being relevant to a user being classified in relation to the selected one of the plurality of specified target features.

9. A computer-implemented method for generating an expanded user segment based on a target segment of users, the method comprising the following steps performed by one or more processors:

receiving information about online activity by a first set of users, the information including a unique user ID associated with each of the first set of users;

extracting, from the received information, data about the first set of users who are associated with a plurality of specified target features;

selecting a combination of two or more of the plurality of specified target features as an output for building at least one model;

building the model that represents the probability of any user being associated with the selected combination of the plurality of specified target features, based on the extracted data about the first set of users;

using the model to score a network population of users relative to the selected combination of the plurality of specified target features, based on information received about online activity by the network population of users; and

generating an expanded user segment by grouping users who have been scored above threshold value for the selected combination of the plurality of specified target features.

10. The computer-implemented method of claim 9, wherein the plurality of specified target features includes one or more of:

a demographic value, a geographic location, a response to a campaign, a conversion, a view of a web page, and a click of an ad.

11. The computer-implemented method of claim 9, wherein the step of building a model is formulated as a supervised machine learning task.

12. The computer-implemented method of claim 9, wherein the model is built based on one or more of:

random forest modeling, linear regression, logistic regression for continuous output values, and Naï

ve Bayesian classification.

13. The computer-implemented method of claim 9, further comprising identifying independent features as being relevant to a user being classified in relation to the selected one of the plurality of specified target features.

14. The computer-implemented method of claim 9, wherein the received information includes one or more of user profiles, cookie data, behavioral logs, leadback logs, click logs, action logs, and impression logs.

15. A system for generating an expanded user segment based on a target segment of users, the system comprising:

a memory storage device configured to store instructions for generating an expanded user segment based on a target segment of users; and

a processor configured to execute the instructions and perform the following steps;

receiving information about online activity by a first plurality of users, the information including a unique user ID associated with each of the first plurality of users;

extracting, from the received information, data about the first plurality of users who are associated with a plurality of specified target features;

selecting a combination of two or more of the plurality of specified target features as an output for building at least one model;

building the model that represents the probability of any user being associated with the selected combination of the plurality of specified target features, based on the extracted data about the first plurality of the users;

using the model to score a second plurality of users relative to the selected combination of the plurality of specified target features, based on information received about online activity by the second plurality of users; and

generating an expanded user segment by grouping users who have been scored above threshold value for the selected combination of the plurality of specified target features.

16. The system of claim 15, wherein the memory includes a large scale storage component and the processor is provided in communication with a distributed caching component.

17. The system of claim 15, wherein the system is disposed in communication with one or more ad servers through the Internet.

18. The system of claim 15, wherein the system is configured to receive cookie data from users through the Internet.