RECOMMENDATION BASED ON LOW-RANK APPROXIMATION

• US 20100325126A1
• Filed: 06/18/2009
• Published: 12/23/2010
• Est. Priority Date: 06/18/2009
• Status: Abandoned Application

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Original Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Inventors: RAJARAM, Shyam S., Scholz, Martin B., Lukose, Rajan, Pan, Rong
• Application Number:
  US12/487,254
• Publication Number:
  US 20100325126A1
• Time in Patent Office: Days
• US Class Current: 707/759
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

A system and method for providing personalized recommendations are disclosed herein. A system includes a processor and a software system executed by the processor. The software system provides a recommendation for an item. The recommendation is based on a comparison of a low-rank approximation of a domain matrix to a user profile. The user profile is based, in part, on the low-rank approximation of the domain matrix.

Claims

1. A personalization system, comprising:

a processor; and

a software system executed by the processor;

wherein the software system provides a recommendation for an item, the recommendation based on a comparison of a low-rank approximation of a domain matrix to a user profile, the user profile based, in part, on the low-rank approximation of the domain matrix.

2. The system of claim 1 wherein the software system computes the low-rank approximation of the domain matrix, the approximation having a pre-selected number of dimensions.

3. The system claim 1, wherein the software system constructs the domain matrix from a domain data set received from a domain storage system, the matrix includes the items in the domain data set and assigns the items to categories to which the items belong as specified by the domain data set.

4. The system of claim 1, wherein the software system constructs the domain matrix as a matrix of binary values, each binary value defining a relation of an item of a domain data set to a category of the domain data set.

5. The system of claim 1, wherein the software system provides an agent for execution on a user computer, the agent provides to the software system information about a preference of the user based on information stored on the user computer and operations performed by the user computer, the preference information corresponds to a domain defined by the domain matrix.

6. The system of claim 5, wherein the agent categorizes the user preference information in accordance with the domain matrix constructed by the software system;

and the agent constructs a user profile based, in part, on the low-rank approximation of the domain matrix.

7. The system of claim 5, wherein the software system receives the user profile from the agent, and determines similarity of a vector of the profile to vectors of the low-rank approximation of the domain matrix.

8. The system of claim 5, wherein the software system categorizes the user preference information in accordance with the domain matrix constructed by the software system, constructs a user profile based, in part, on the low-rank approximation of the domain matrix, and determines similarity of a vector of the profile vectors to vectors of the low-rank approximation of the domain matrix.

9. A computer readable medium encoded with a computer program, the computer program comprising:

instructions that when executed by a processor compute a low-rank approximation of a matrix representing a domain data set, the approximation matrix having a pre-selected number of dimensions; and

instructions that when executed by a processor provide a recommendation for an item, the recommendation based on similarity of the low-rank approximation of the matrix representing the domain data set to a user profile based, in part, on the low-rank approximation of the matrix representing the domain data set.

10. The computer readable medium of claim 9, further comprising instructions that when executed by a processor construct the matrix representing the domain data set, the matrix includes items in the domain data set and assigns the items to domain categories to which the items belong as specified by the domain data set.

11. The computer readable medium of claim 9, further comprising instructions that when executed by a processor construct the matrix representing the domain data set as a matrix of binary values, each binary value defining a relation of an item of the domain data set to a category of the domain data set.

12. The computer readable medium of claim 9, further comprising instructions that when executed by a processor provide an agent for execution on a user computer, the agent provides information about a preference of a user based on information stored on the user computer and operations performed by the user computer.

13. The computer readable medium of claim 9, further comprising:

instructions that when executed by a processor categorize user preference information in accordance with the matrix representing the domain data set; and

instructions that when executed by a processor construct a user profile based, in part, on the categorized user preference information and the low-rank approximation of the matrix representing the domain data set.

14. The computer readable medium of claim 9, further comprising instructions that when executed by a processor receive, from a user computer, a user profile based on the low-rank approximation of the matrix representing the domain data set, and determine similarity of a vector of the profile to vectors of the low-rank approximation of the domain matrix.

15. The computer readable medium of claim 9, further comprising:

instructions that when executed by a processor generate a profile for an item of the domain data set, the profile based on the low-rank approximation of the matrix representing the domain data set; and

instructions that when executed by a processor generate a profile for an item comprising a plurality of items of the domain data set, the profile based on the low-rank approximation of the matrix representing the domain data set.

16. A method, comprising:

computing, by a processor, a low-rank approximation of a matrix representing a domain data set, the approximation matrix having a pre-selected number of dimensions;

providing a recommendation for an item, via a processor, the recommendation based on a comparison of the low-rank approximation of the matrix representing the domain data set to a user profile based, in part, on the low-rank approximation of the matrix representing the domain data set.

17. The method of claim 16, further comprising:

selecting the domain data set, wherein the domain data set represents items and relationships between items in a selected domain;

constructing, by a processor, the matrix representing the domain data set, the matrix includes the items in the domain data set and assigns the items to categories to which the items belong as specified by the domain data set; and

selecting a number of dimensions for the low-rank approximation of the matrix representing the domain data set.

18. The method of claim 17, further comprising constructing the matrix representing the domain data set as a matrix of binary values, each binary value defining a relation of an item of the domain data set to a category of the domain data set.

19. The method of claim 16, further comprising providing an agent for execution on a user computer, the agent provides information about a preference of a user based on information stored on the user computer and operations performed by the user computer.

20. The method of claim 16, further comprising:

categorizing user preference information in accordance with the matrix representing the domain data set; and

constructing a user profile based, in part, on the categorized user preference information and the low-rank approximation of the matrix representing the domain data set.