Privacy-aware on-line user role tracking

• US 8,775,335 B2
• Filed: 08/05/2011
• Issued: 07/08/2014
• Est. Priority Date: 08/05/2011
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Lin, Ching-Yung, Tong, Hanghang, Wang, Fei
• Primary Examiner(s): BURKE, JEFF A
• Application Number:
  US13/198,790
• Publication Number:
  US 20130036116A1
• Time in Patent Office: 1,068 Days
• Field of Search: None
• US Class Current: 706/12
• CPC Class Codes: G06F 18/2135 based on approximation criteria, e.g. principal component analysis
• CPC Class Codes: G06F 2218/08 Feature extraction
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

Access is obtained to a first nonnegative factor matrix and a second nonnegative factor matrix obtained by factorizing a nonnegative asymmetric matrix which represents a set of data which tracks time-stamped activities of a plurality of entities. The first nonnegative factor matrix is representative of initial role membership of the entities, and the second nonnegative factor matrix is representative of initial role activity descriptions. At a given one of the time stamps, while holding a change in the first nonnegative factor matrix constant, a change in the second nonnegative factor matrix is updated to reflect time variance of the set of data at the given one of the time stamps, without accessing actual data values at previous ones of the time stamps. At the given one of the time stamps, while holding a change in the second nonnegative factor matrix constant, a change in the first nonnegative factor matrix is updated, to reflect the time variance of the set of data at the given one of the time stamps, without accessing the actual data values at the previous ones of the time stamps. The role membership of the entities and the role activity descriptions, at the given one of the time stamps, are updated based on the updating steps. A suitable technique for nonnegative symmetric matrices is also provided.

Claims

1. A method for analyzing a set of data which tracks time-stamped activities of a plurality of entities, said set of data comprising a nonnegative asymmetric matrix, said method comprising:

obtaining access to a first nonnegative factor matrix and a second nonnegative factor matrix obtained by factorizing said nonnegative asymmetric matrix, said first nonnegative factor matrix being representative of initial role membership of said entities, said second nonnegative factor matrix being representative of initial role activity descriptions;

at a given one of said time stamps, while holding a change in said first nonnegative factor matrix constant, updating a change in said second nonnegative factor matrix, to reflect time variance of said set of data at said given one of said time stamps, without accessing actual data values at previous ones of said time stamps, said actual data values at said previous ones of said time stamps being privacy-sensitive;

at said given one of said time stamps, while holding a change in said second nonnegative factor matrix constant, updating a change in said first nonnegative factor matrix, to reflect said time variance of said set of data at said given one of said time stamps, without accessing said actual data values at said previous ones of said time stamps, said actual data values at said previous ones of said time stamps being privacy-sensitive;

updating said role membership of said entities and said role activity descriptions, at said given one of said time stamps, based on said updating steps; and

providing a system, wherein the system comprises distinct software modules, each of the distinct software modules being embodied on a non-transitory computer-readable storage medium, and wherein the distinct software modules comprise an initialization module, an inference module, and an update module;

wherein;

said obtaining step is carried out by said initialization module executing on at least one hardware processor;

said updating of said change in said second nonnegative factor matrix and said updating of said change in said first nonnegative factor matrix are carried out by said inference module executing on said at least one hardware processor; and

said updating of said role membership of said entities and said role activity descriptions is carried out by said update module executing on said at least one hardware processor.

2. The method of claim 1, wherein said steps of updating said change in said second nonnegative factor matrix and said change in said first nonnegative factor matrix comprise use of a projected gradient technique.

3. The method of claim 2, wherein said updating of said change in said second nonnegative factor matrix comprises:

calculating interaction between a change of entity-role membership and an old role-activity description;

calculating an interaction between an old entity-role membership and a change of role-activity description;

calculating an interaction between said change of entity-role membership and said change of role-activity description;

inferring strength of said change of role-activities description; and

adjusting said strength of said change of role-activities description to feasible regions.

4. The method of claim 3, wherein said updating of said change in said first nonnegative factor matrix comprises:

calculating said interaction between said change of entity-role membership and said old role-activity description;

calculating said interaction between said old entity-role membership and said change of role-activity description;

calculating an interaction between said change of entity-role membership and said change of role-activity description;

inferring strength of said change of entity-role membership; and

adjusting said strength of said change of entity-role membership to feasible regions.

5. The method of claim 2, further comprising repeating said steps of updating said change in said second nonnegative factor matrix and said change in said first nonnegative factor matrix, and said step of updating said role membership of said entities and said role activity descriptions, at a plurality of additional ones of said time stamps.

6. A method for analyzing a set of data which tracks time-stamped activities of a plurality of entities, said set of data comprising a nonnegative symmetric matrix, said method comprising:

obtaining access to a nonnegative factor matrix, obtained by factorizing said nonnegative symmetric matrix, said nonnegative factor matrix being representative of initial role membership of said entities;

at a given one of said time stamps, updating a change in said nonnegative factor matrix, to reflect time variance of said set of data at said given one of said time stamps, without accessing actual data values at previous ones of said time stamps, said actual data values at said previous ones of said time stamps being privacy-sensitive;

updating said role membership of said entities, at said given one of said time stamps, based on said updating step; and

providing a system, wherein the system comprises distinct software modules, each of the distinct software modules being embodied on a non-transitory computer-readable storage medium, and wherein the distinct software modules comprise an initialization module, an inference module, and an update module;

wherein;

said obtaining step is carried out by said initialization module executing on at least one hardware processor;

said updating of said change in said nonnegative factor matrix is carried out by said inference module executing on said at least one hardware processor; and

said updating of said role membership of said entities is carried out by said update module executing on said at least one hardware processor.

7. The method of claim 6, wherein said step of updating said change in said nonnegative factor matrix comprises use of a projected gradient technique.

8. The method of claim 7, wherein said updating of said change in said nonnegative factor matrix comprises:

calculating interaction between a change of entity-role membership and an old entity-role membership;

inferring strength of the change of entity-role membership; and

adjusting said strength of said change of entity-role membership to feasible regions.

9. The method of claim 7, further comprising repeating said step of updating said change in said nonnegative factor matrix and said step of updating said role membership of said entities at a plurality of additional ones of said time stamps.

10. An article of manufacture comprising a computer program product for analyzing a set of data which tracks time-stamped activities of a plurality of entities, said set of data comprising a nonnegative asymmetric matrix, said computer program product comprising:

a non-transitory tangible computer-readable storage medium, storing in a non-transitory manner computer readable program code, the computer readable program code comprising;

computer readable program code configured to, when executed on at least one hardware processor, cause said at least one hardware processor to obtain access to a first nonnegative factor matrix and a second nonnegative factor matrix obtained by factorizing said nonnegative asymmetric matrix, said first nonnegative factor matrix being representative of initial role membership of said entities, said second nonnegative factor matrix being representative of initial role activity descriptions;

computer readable program code configured to, when executed on said at least one hardware processor, cause said at least one hardware processor to, at a given one of said time stamps, while holding a change in said first nonnegative factor matrix constant, update a change in said second nonnegative factor matrix, to reflect time variance of said set of data at said given one of said time stamps, without accessing actual data values at previous ones of said time stamps, said actual data values at said previous ones of said time stamps being privacy-sensitive;

computer readable program code configured to, when executed on said at least one hardware processor, cause said at least one hardware processor to, at said given one of said time stamps, while holding a change in said second nonnegative factor matrix constant, update a change in said first nonnegative factor matrix, to reflect said time variance of said set of data at said given one of said time stamps, without accessing said actual data values at said previous ones of said time stamps, said actual data values at said previous ones of said time stamps being privacy-sensitive; and

computer readable program code configured to, when executed on said at least one hardware processor, cause said at least one hardware processor to update said role membership of said entities and said role activity descriptions, at said given one of said time stamps, based on said updating of said changes in said first and second nonnegative factor matrices;

wherein the computer readable program code comprises distinct software modules, each of the distinct software modules being embodied on said non-transitory tangible computer-readable storage medium, and wherein the distinct software modules comprise an initialization module, an inference module, and an update module;

wherein;

said initialization module comprises said computer readable program code configured to obtain access

said inference module comprises said computer readable program code configured to update said change in said second nonnegative factor matrix and to update said change in said first nonnegative factor matrix; and

said update module comprises said computer readable program code configured to update said role membership of said entities and said role activity descriptions.

11. The article of manufacture of claim 10, wherein said computer readable program code configured to update said change in said second nonnegative factor matrix and said change in said first nonnegative factor matrix comprises computer readable program code configured to use a projected gradient technique.

12. The article of manufacture of claim 11, wherein said computer readable program code configured to update said change in said second nonnegative factor matrix comprises:

computer readable program code configured to calculate interaction between a change of entity-role membership and an old role-activity description;

computer readable program code configured to calculate interaction between an old entity-role membership and a change of role-activity description;

computer readable program code configured to calculate interaction between said change of entity-role membership and said change of role-activity description;

computer readable program code configured to infer strength of said change of role-activities description; and

computer readable program code configured to adjust said strength of said change of role-activities description to feasible regions.

13. The article of manufacture of claim 12, wherein said computer readable program code configured to update said change in said first nonnegative factor matrix comprises:

computer readable program code configured to calculate said interaction between said change of entity-role membership and said old role-activity description;

computer readable program code configured to calculate said interaction between said old entity-role membership and said change of role-activity description;

computer readable program code configured to calculate an interaction between said change of entity-role membership and said change of role-activity description;

computer readable program code configured to infer strength of said change of entity-role membership; and

computer readable program code configured to adjust said strength of said change of entity-role membership to feasible regions.

14. The article of manufacture of claim 11, further comprising computer readable program code configured to repeat:

said updating of said change in said second nonnegative factor matrix and said updating of said change in said first nonnegative factor matrix, and

said updating said role membership of said entities and said role activity descriptions, at a plurality of additional ones of said time stamps.

15. An article of manufacture comprising a computer program product for analyzing a set of data which tracks time-stamped activities of a plurality of entities, said set of data comprising a nonnegative symmetric matrix, said computer program product comprising:

a non-transitory tangible computer-readable storage medium, storing in a non-transitory manner computer readable program code, the computer readable program code comprising;

computer readable program code configured to, when executed on at least one hardware processor, cause said at least one hardware processor to obtain access to a nonnegative factor matrix, obtained by factorizing said nonnegative symmetric matrix, said nonnegative factor matrix being representative of initial role membership of said entities;

computer readable program code configured to, when executed on at least one hardware processor, cause said at least one hardware processor to, at a given one of said time stamps, update a change in said nonnegative factor matrix, to reflect time variance of said set of data at said given one of said time stamps, without accessing actual data values at previous ones of said time stamps, said actual data values at said previous ones of said time stamps being privacy-sensitive; and

computer readable program code configured to, when executed on at least one hardware processor, cause said at least one hardware processor to update said role membership of said entities, at said given one of said time stamps, based on said updating of said change in said nonnegative factor matrix;

wherein the computer readable program code comprises distinct software modules, each of the distinct software modules being embodied on said non-transitory tangible computer-readable storage medium, and wherein the distinct software modules comprise an initialization module, an inference module, and an update module;

wherein;

said initialization module comprises said computer readable program code configured to obtain access

said inference module comprises said computer readable program code configured to update said change in said nonnegative factor matrix; and

said update module comprises said computer readable program code configured to update said role membership of said entities.

16. The article of manufacture of claim 15, wherein said step of updating said change in said nonnegative factor matrix comprises use of a projected gradient technique.

17. The article of manufacture of claim 16, wherein said computer readable program code configured to update said change in said nonnegative factor matrix comprises:

computer readable program code configured to calculate interaction between a change of entity-role membership and an old entity-role membership;

computer readable program code configured to infer strength of the change of entity-role membership; and

computer readable program code configured to adjust said strength of said change of entity-role membership to feasible regions.

18. The article of manufacture of claim 16, further comprising computer readable program code configured to repeat said updating of said change in said nonnegative factor matrix and said updating of said role membership of said entities at a plurality of additional ones of said time stamps.

19. An apparatus for analyzing a set of data which tracks time-stamped activities of a plurality of entities, said set of data comprising a nonnegative asymmetric matrix, said apparatus comprising:

a memory;

at least one processor coupled to said memory; and

a persistent storage device coupled to said memory and said at least one processor;

wherein said persistent storage device comprises machine-executable code which when loaded into said memory causes said at least one processor to be operative to;

obtain access to a first nonnegative factor matrix and a second nonnegative factor matrix obtained by factorizing said nonnegative asymmetric matrix, said first nonnegative factor matrix being representative of initial role membership of said entities, said second nonnegative factor matrix being representative of initial role activity descriptions;

at a given one of said time stamps, while holding a change in said first nonnegative factor matrix constant, update a change in said second nonnegative factor matrix, to reflect time variance of said set of data at said given one of said time stamps, without accessing actual data values at previous ones of said time stamps, said actual data values at said previous ones of said time stamps being privacy-sensitive;

at said given one of said time stamps, while holding a change in said second nonnegative factor matrix constant, update a change in said first nonnegative factor matrix, to reflect said time variance of said set of data at said given one of said time stamps, without accessing said actual data values at said previous ones of said time stamps, said actual data values at said previous ones of said time stamps being privacy-sensitive; and

update said role membership of said entities and said role activity descriptions, at said given one of said time stamps, based on said updating of said first and second nonnegative factor matrices;

wherein said persistent storage device comprises a computer-readable storage medium, further comprising a plurality of distinct software modules, each of the distinct software modules being embodied on said computer-readable storage medium, and wherein the distinct software modules comprise an initialization module, an inference module, and an update module; and

wherein;

said at least one processor is operative to obtain access by executing said initialization module;

said at least one processor is operative to update said change in said second nonnegative factor matrix and to update said change in said first nonnegative factor matrix by executing said inference module; and

said at least one processor is operative to update said role membership of said entities and said role activity descriptions by executing said update module.

20. The apparatus of claim 19, wherein said at least one processor is operative to update said change in said second nonnegative factor matrix and said change in said first nonnegative factor matrix by use of a projected gradient technique.

21. The apparatus of claim 20, wherein said at least one processor is operative to update said change in said second nonnegative factor matrix by:

calculating interaction between a change of entity-role membership and an old role-activity description;

calculating an interaction between an old entity-role membership and a change of role-activity description;

calculating an interaction between said change of entity-role membership and said change of role-activity description;

inferring strength of said change of role-activities description; and

adjusting said strength of said change of role-activities description to feasible regions.

22. The apparatus of claim 21, wherein said at least one processor is operative to update said change in said first nonnegative factor matrix by:

calculating said interaction between said change of entity-role membership and said old role-activity description;

calculating said interaction between said old entity-role membership and said change of role-activity description;

calculating an interaction between said change of entity-role membership and said change of role-activity description;

inferring strength of said change of entity-role membership; and

adjusting said strength of said change of entity-role membership to feasible regions.