UNIVERSAL SYSTEM AND METHOD FOR REPRESENTING AND PREDICTING HUMAN BEHAVIOR

• US 20090248599A1
• Filed: 03/31/2009
• Published: 10/01/2009
• Est. Priority Date: 04/01/2008
• Status: Active Grant

• Current Assignee: Certona Corporation (StartApp, Inc.)
• Original Assignee: Geoffrey J. Hueter, Noble H. Hueter, Steven C. Quandt
• Inventors: Hueter, Geoffrey J., Hueter, Noble H., Quandt, Steven C.
• Granted Patent: US 8,566,256 B2
• Time in Patent Office: Days
• US Class Current: 706/20
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06Q 10/00 Administration; Management
• CPC Class Codes: G06Q 30/00 Commerce
• CPC Class Codes: G06Q 30/0185 Product, service or business identity fraud
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06Q 30/0204 Market segmentation
• CPC Class Codes: G06Q 30/0255 based on user history
• CPC Class Codes: G06Q 30/0269 based on user profile or attribute
• CPC Class Codes: G06Q 30/0631 Item recommendations
• CPC Class Codes: H04L 51/212 using filtering or selective blocking

Abstract

A system and method is disclosed for profiling subjects and objects based on subjects'"'"' responses to various objects for purposes of determining and presenting the objects most likely to generate the most positive response from each visitor. Object ratings, such as aesthetic response, preference, interest, or relevancy, are explicitly submitted by subjects or derived implicitly from visitor interactions with the objects. Objects include movies, books, songs, commercial products, news articles, advertisements or any other type of content or physical item. A profiling engine processes the ratings information and generates compact profiles of each subject and object based on the similarities and differences in affinities between the group of subjects and the group of objects. A recommendation engine then generates recommendations to a subject based on similarity between the subject and object profiles. The recommendation engine can also match subjects to other subjects and objects to other objects. The recommendation engine can also predict affinity across object catalogs and across time. Additionally, the object profiles can be clustered to create behavioral object categories. The system has application in personalization, behavioral targeting, Internet retailing and interactive radio, to name but a few applications.

Claims

1. A universal system for representing and predicting human behavior, comprising:

(a) a service system to collect ratings and provide recommendations;

(b) subjects and objects represented in vector form;

(c) optimal subject and object vectors derived solely from said subject'"'"'s ratings of objects;

(d) recommendations to, and collecting ratings from, an external application; and

(e) piecewise separable calculation method for deterministic and distributable processing; and

whereby said system does not need to know anything about object content, does not need to know anything about subject demographics, employs statistical/neural network modeling by means of input modeling rather than transfer function, uses a fixed size profile for subjects and objects for scalable processing, and

all accomplished in a non-invasive, mentor-less and self-optimizing fashion, scalable to large numbers of users and objects.

2. The universal system for representing and predicting human behavior, according to claim 1, wherein said service system to collect ratings and provide recommendations includes collecting ratings that are implicit or explicit.

3. The universal system for representing and predicting human behavior, according to claim 1, wherein said subjects and objects represented in vector form includes subjects and objects profiled at different times and matched across time.

4. The universal system for representing and predicting human behavior, according to claim 1, wherein said subjects and objects represented in vector form includes similar subjects or similar objects that are clustered together in vector space, and further wherein said clusters can be aligned across separate applications'"'"' vector spaces.

5. The universal system for representing and predicting human behavior, according to claim 4, wherein the variety of recommendations can be maximized by dispersing said recommendations across a targeted cluster.

6. The universal system for representing and predicting human behavior, according to claim 1, wherein said subjects and objects represented in vector form includes vectors having dimensions derived from an analysis of the ratings data and not requiring a pre-defined set of attributes.

7. The universal system for representing and predicting human behavior, according to claim 6, wherein said vectors derived from an analysis of the ratings data and not requiring a pre-defined set of attributes includes optimal subject and object vectors derived solely from said subject'"'"'s ratings of objects includes deriving vectors by a factorization process which minimizes the prediction error of the ratings from the matching of the derived vectors and which self-determines the number of dimensions of the vectors, both individually and collectively, as well as employs progressive training with self-determination of the number of dimensions used.

8. The universal system for representing and predicting human behavior, according to claim 1, wherein said external application includes a web page, a personalized radio, a personal computer, an in-store kiosk, a personal digital assistant (PDA), a mobile phone, a set top box, or a conventional telephone.

9. The universal system for representing and predicting human behavior, according to claim 1, wherein said recommendations includes recommendations that are made to a self-associated group.

10. The universal system for representing and predicting human behavior, according to claim 1, wherein said external application includes the said system to profile and make recommendations to any type of individually identified objects managed and presented by the external application.

11. The universal system for representing and predicting human behavior, according to claim 10, wherein said individually identified objects include consumer products, music, books, theatre, art, employer/employee matching, and travel.

12. The universal system for representing and predicting human behavior, according to claim 1, wherein said piecewise separable calculation method includes an iterative calculation method wherein an individual subject vector can be updated from the existing set of object vectors and an individual object vector can be updated from the existing set of subject vectors, and said calculation is repeated until a specified convergence level is reached.

13. A universal method for representing and predicting human behavior, comprising the steps of:

(a) providing a service to collect ratings and provide recommendations;

(b) representing subjects and objects in vector form;

(c) deriving optimal subject and object vectors solely from subject'"'"'s ratings of objects;

(d) providing recommendations to, and collecting ratings from, an external application; and

(e) providing piecewise separable calculation method for deterministic and distributable processing; and

whereby said system does not need to know anything about object content, does not need to know anything about subject demographics, employs statistical/neural network modeling by means of input modeling rather than transfer function, uses a fixed size profile for subjects and objects for scalable processing, and

all accomplished in a non-invasive, mentor-less and self-optimizing fashion, scalable to large numbers of users and objects.

14. The universal method for representing and predicting human behavior, according to claim 13, wherein said step of providing a service to collect ratings and provide recommendations includes the step of providing a service system to collect ratings and provide recommendations includes collecting ratings that are implicit or explicit.

15. The universal method for representing and predicting human behavior, according to claim 13, wherein said step of representing subjects and objects in vector form includes the step of representing subjects and objects in vector form wherein subjects and objects represented in vector form includes subjects and objects profiled at different times and matched across time.

16. The universal method for representing and predicting human behavior, according to claim 13, wherein said step of representing subjects and objects in vector form includes the step of representing subjects and objects in vector form wherein similar subjects or similar objects that are clustered together in vector space, and further wherein said clusters can be aligned across separate applications'"'"' vector spaces.

17. The universal method for representing and predicting human behavior, according to claim 16, wherein said step of providing recommendations includes the step of providing recommendations wherein the variety of recommendations can be maximized by dispersing said recommendations across a targeted cluster.

18. The universal method for representing and predicting human behavior, according to claim 13, wherein said step of representing subjects and objects in vector form includes the step of representing subjects and objects in vector form wherein said vectors have dimensions derived from an analysis of the ratings data and not requiring a pre-defined set of attributes.

19. The universal method for representing and predicting human behavior, according to claim 18, wherein said step of deriving optimal subject and object vectors solely from subject'"'"'s ratings of objects includes the step of deriving vectors from an analysis of the ratings data and not requiring a pre-defined set of attributes includes derived optimal subject and object vectors solely from subject'"'"'s ratings of objects by a factorization process which minimizes the prediction error of the ratings from the matching of the derived vectors and which self-determines the number of dimensions of the vectors, both individually and collectively.

20. The universal method for representing and predicting human behavior, according to claim 13, wherein said step of providing recommendations to, and collecting ratings from, an external application includes the step of providing recommendations to, and collecting ratings from, an external application where said external application includes a web page, a personalized radio, a personal computer, an in-store kiosk, a personal digital assistant (PDA), a mobile phone, a set top box, or a conventional telephone.

21. The universal method for representing and predicting human behavior, according to claim 13, wherein said step of providing piecewise separable calculation method for deterministic and distributable processing includes the step of providing piecewise separable calculation method for deterministic and distributable processing comprising an iterative calculation method wherein an individual subject vector can be updated from the existing set of object vectors and an individual object vector can be updated from the existing set of subject vectors, and said calculation is repeated until a specified convergence level is reached.

22. The universal method for representing and predicting human behavior, according to claim 13, wherein said step of providing recommendations to, and collecting ratings from, an external application includes the step of providing recommendations to, and collecting ratings from, an external application wherein said external application includes the said system to profile and make recommendations of any type of individually identified objects managed and presented by the external application, and further wherein said individually identified objects include consumer products, music, books, theatre, art, employer/employee matching, and travel.

23. A method for using a universal system for representing and predicting human behavior, comprising the steps of:

(a) collecting individual subject object ratings;

(b) storing said subject object ratings;

(c) generating subject and object vectors from said stored subject object ratings;

(d) storing said subject and object vectors; and

(e) scoring said stored subject and object vectors against other said subject and object vectors to generate recommendations.

24. The universal method for using a system for representing and predicting human behavior, according to claim 23, wherein said step of scoring said stored subject and object vectors against said subject and object vectors to generate recommendations includes the step of scoring said stored subject and object vectors against said subject and object vectors to generate recommendations within a personalized radio application.