Inferring profiles of network users and the resources they access

• US 8,494,897 B1
• Filed: 06/29/2009
• Issued: 07/23/2013
• Est. Priority Date: 06/30/2008
• Status: Active Grant

• Current Assignee: Alexa Internet (Amazon.com, Inc.)
• Original Assignee: Alexa Internet (Amazon.com, Inc.)
• Inventors: Dawson, Steven
• Primary Examiner(s): Rines, David
• Application Number:
  US12/494,151
• Time in Patent Office: 1,485 Days
• Field of Search: 705 711- 742, 705 147- 1473, 709201-253
• US Class Current: 705/7.33
• CPC Class Codes: G06Q 10/10 Office automation; Time management
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

An analytics system and associated methods are disclosed. The analytics system may be used by site owners and others for various purposes, such as to identify and compare network sites that satisfy particular criteria, to identify and compare keywords associated with these sites, and to assess characteristics of typical users of these sites. One feature of the system includes computer processes for inferring characteristics of users for whom user profile data is unknown or incomplete, and for using these inferred characteristics to more accurately assess the characteristics of typical users of particular sites.

Claims

1. A method of generating demographic profiles of network-accessible sites, the method comprising:

obtaining demographic information relating to each network user of a first set of network users, said demographic data supplied by the respective users;

obtaining network usage information relating to each of the network users of the first set of network users;

training a computer-implemented classifier using the demographic information and the network usage information relating to each of the network users of the first set;

obtaining network usage information for a second set of users for whom demographic information is unknown or incomplete;

applying the computer-implemented classifier to the network usage information of said second set of users to generate predictive demographic information for the users in the second set; and

generating, by a computer system that comprises one or more computing devices, a demographic profile of a network-accessible site based on a combination of (1) the demographic information supplied by users who have accessed the site, and (2) the predictive demographic information of users who have accessed the site, wherein generating the demographic profile of the site comprises giving an amount of weight to a user'"'"'s supplied or predictive demographic information that is proportional to a measure of the user'"'"'s quantity of usage of the site.

2. The method of claim 1, wherein training the classifier comprises using at least some of the network usage information relating to each of the network users of the first set as an input training set, and using at least some of the demographic information relating to each of the network users of the first set as an output training set.

3. The method of claim 1, wherein the network usage information relating to each of the network users of the first set comprises search query submissions of users.

4. The method of claim 1, wherein generating the demographic profile of the site comprises:

generating a first demographic profile of the site based on the supplied demographic information of users who have accessed the site;

generating a second demographic profile of the site based on the predictive demographic information of users who have accessed the site; and

combining the first and second demographic profiles.

5. The method of claim 4, wherein combining the first and second demographic profiles comprises accounting for differences between (1) categories used to classify users who have supplied demographic information, and (2) categories used to classify users who have not supplied demographic information.

6. The method of claim 1, wherein obtaining the network usage information comprises obtaining usage trail information from a client software program executed by computing devices of the first set of network users.

7. The method of claim 6, wherein the client software program is a browser plugin.

8. The method of claim 1, wherein obtaining the network usage information comprises obtaining network usage information using cookies stored on computing devices of users.

9. The method of claim 1, wherein the classifier is a voting classifier comprising at least two sub-classifiers, wherein each sub-classifier is selected from the group consisting of a naï

ve Bayes classifier, a decision tree, an artificial neural network, a support vector machine, an expert system, and a nearest neighbor classifier.

10. A method of generating demographic profiles of network-accessible sites, the method comprising:

obtaining demographic information relating to each network user of a first set of network users, said demographic data supplied by the respective users;

obtaining network usage information relating to each of the network users of the first set of network users;

training a computer-implemented classifier using the demographic information and the network usage information relating to each of the network users of the first set;

obtaining network usage information for a second set of users for whom demographic information is unknown or incomplete;

applying the computer-implemented classifier to the network usage information of said second set of users to generate predictive demographic information for the users in the second set;

generating, by a computer system that comprises one or more computing devices, a demographic profile of a network-accessible site based on a combination of (1) the demographic information supplied by users who have accessed the site, and (2) the predictive demographic information of users who have accessed the site;

using the demographic profile of the site, in combination with demographic profiles generated for other sites, to provide an interactive site screener that enables an operator, via a demographic filter user interface, to specify multiple demographic characteristics and to obtain a corresponding listing of sites;

receiving, from a user of the interactive site screener, a demographic filter specified via the demographic filter user interface, said demographic filter specifying a plurality of demographic characteristics selected from a plurality of predefined demographic categories of the interactive site screener; and

generating a listing of sites for presentation to the user of the interactive site screener based on the received demographic filter and the demographic profiles of the sites.

11. The method of claim 10, wherein the interactive site screener is configured to identify sites for which more than a threshold percentage of visitors satisfy operator-specified demographic criteria.

12. Non-transitory computer storage that stores executable program code that directs a computing system comprising one or more computers to generate demographic profiles of network-accessible sites by a process that comprises:

obtaining demographic information of a first set of users, said demographic information supplied by respective users;

obtaining network usage information reflective of network access behaviors of the first set of users;

training a classifier using the demographic information and the network usage information associated with the first set of users;

obtaining network usage information for a second set of users for whom demographic information is unknown or incomplete;

applying the classifier to the network usage information of the second set of users to generate inferred demographic information for the users in the second set; and

generating a demographic profile of a network-accessible site using both (1) the demographic information supplied by users who have accessed the site, and (2) the inferred demographic information of users who have accessed the site, wherein generating the demographic profile of the site comprises giving an amount of weight to a user'"'"'s supplied or inferred demographic information that is proportional to a measure of the user'"'"'s quantity of usage of the site.

13. The non-transitory computer storage of claim 12, wherein training the classifier comprises using at least some of the network usage information of users of the first set as an input training set, and using at least some of the demographic information of users of the first set as an output training set.

14. The non-transitory computer storage of claim 12, wherein generating the demographic profile of the site comprises:

generating a first demographic profile of the site based on the supplied demographic information of users who have accessed the site;

generating a second demographic profile of the site based on the inferred demographic information of users who have accessed the site; and

combining the first and second demographic profiles.

15. The non-transitory computer storage of claim 14, wherein combining the first and second profiles comprises accounting for differences between (1) categories used to classify users who have supplied demographic information and (2) categories used to classify users who have not supplied demographic information.

16. The non-transitory computer storage of claim 12, wherein the classifier is a voting classifier comprising at least two sub-classifiers.

17. The non-transitory computer storage of claim 12, wherein the classifier generates likelihood information representing a likelihood that a target user has a particular demographic characteristic given the existence of a particular network usage characteristic of the target user.

18. The non-transitory computer storage of claim 12, wherein the first set of users comprises toolbar users who have explicitly supplied their demographic information, and the second set of users includes toolbar users who have not supplied their demographic information.

19. The non-transitory computer storage of claim 12, in combination with the computing system, wherein the computing system is programmed with the executable program code to implement the process.

20. Non-transitory computer storage that stores executable program code that directs a computing system comprising one or more computers to generate demographic profiles of network-accessible sites by a process that comprises:

obtaining demographic information of a first set of users, said demographic information supplied by respective users;

obtaining network usage information reflective of network access behaviors of the first set of users;

training a classifier using the demographic information and the network usage information associated with the first set of users;

obtaining network usage information for a second set of users for whom demographic information is unknown or incomplete;

applying the classifier to the network usage information of the second set of users to generate inferred demographic information for the users in the second set; and

generating a demographic profile of a network-accessible site using both (1) the demographic information supplied by users who have accessed the site, and (2) the inferred demographic information of users who have accessed the site;

wherein the executable program code further directs the computing system to use the demographic profile of the site, in combination with demographic profiles generated for other sites, to provide an interactive site screener that enables an operator, via a demographic filter user interface, to specify multiple demographic characteristics and to obtain a corresponding listing of sites;

wherein the executable program code is responsive to a submission by a site screener user of a demographic filter by directing the computing system to generate, based on the demographic profiles of the sites, a listing of sites corresponding to the demographic filter, said demographic filter comprising a plurality of demographic characteristics selected from a plurality of predefined demographic categories of the interactive site screener.

21. The non-transitory computer storage of claim 20, wherein the interactive site screener is configured to identify sites for which more than a selected percentage of visitors satisfy operator-specified demographic criteria.

22. A system, comprising:

a computer data repository that stores network usage data of a first set of users who have supplied demographic information and for a second set of users who have not supplied demographic information, the network usage data reflecting network usage behaviors, including site visits, of particular users;

a computer system comprising one or more computers, said computer system programmed via executable instructions to at least;

generate inferred demographic information for users in the second set based on the network usage data of the users in the first and second sets and the demographic information supplied by users in the first set; and

generate a demographic profile of a network-accessible site based on a combination of (1) the supplied demographic information of users in the first set who have accessed the site, and (2) the inferred demographic information of users in the second set who have accessed the site;

wherein the computer system is programmed to generate the demographic profile of the site by a process that comprises giving an amount of weight to a user'"'"'s supplied or inferred demographic information that is proportional to a measure of the user'"'"'s quantity of usage of the site.

23. The system of claim 22, wherein the computer system is programmed to generate the demographic profile of the site by a process that comprises:

generating a first demographic profile of the site based on the supplied demographic information of users in the first set who have accessed the site;

generating a second demographic profile of the site based on the inferred demographic information of users in the second set who have accessed the site; and

combining the first and second demographic profiles.

24. A system, comprising:

a computer data repository that stores network usage data of a first set of users who have supplied demographic information and for a second set of users who have not supplied demographic information, the network usage data reflecting network usage behaviors, including site visits, of particular users;

a computer system comprising one or more computers, said computer system programmed via executable instructions to at least;

generate inferred demographic information for users in the second set based on the network usage data of the users in the first and second sets and the demographic information supplied by users in the first set;

generate a demographic profile of a network-accessible site based on a combination of (1) the supplied demographic information of users in the first set who have accessed the site, and (2) the inferred demographic information of users in the second set who have accessed the site;

use the demographic profile of the site, in combination with demographic profiles generated for other sites, to provide an interactive site screener that enables an operator, via a demographic filter user interface, to specify multiple demographic characteristics and to obtain a corresponding listing of sites;

receive, from a user of the interactive site screener, a demographic filter that includes a plurality of demographic criteria selected from a plurality of demographic categories presented on the demographic filter user interface; and

generate, based on the demographic filter and demographic profiles of sites, a listing of sites to present to the user.

25. The system of claim 24, wherein the interactive site screener is configured to identify sites for which more than a selected percentage of visitors satisfy operator-specified demographic criteria.

26. The system of claim 24, wherein the demographic filter user interface includes user functionality for selecting demographic criteria from at least two of the following demographic categories:

country, ethnicity, education level, income level.

27. The system of claim 24, wherein the computer system is additionally programmed to generate, for a particular site, a graphical indication of how a demographic profile of the particular site differs from an average demographic profile taken over multiple sites.