Non-negative matrix factorization in a relational database management system

• US 20050246354A1
• Filed: 08/27/2004
• Published: 11/03/2005
• Est. Priority Date: 08/29/2003
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Milenova, Boriana L., McCracken, Mark A., Campos, Marcos M., Yarmus, Joseph S., Tamayo, Pablo, Tang, George G., Jagannath, Mahesh K., Kuntala, Pavani
• Granted Patent: US 7,734,652 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/2465 Query processing support for facilitating data mining operations in structured databases
• CPC Class Codes: G06F 16/283 Multi-dimensional databases or data warehouses, e.g. MOLAP or ROLAP

Abstract

An implementation of NMF functionality integrated into a relational database management system provides the capability to apply NMF to relational datasets and to sparse datasets. A database management system comprises a multi-dimensional data table operable to store data and a processing unit operable to perform non-negative matrix factorization on data stored in the multi-dimensional data table and to generate a plurality of data tables, each data table being smaller than the multi-dimensional data table and having reduced dimensionality relative to the multi-dimensional data table. The multi-dimensional data table may be a relational data table.

Claims

1. A database management system comprising:

a multi-dimensional data table operable to store data; and

a processing unit operable to perform non-negative matrix factorization on data stored in the multi-dimensional data table and to generate a plurality of data tables, each data table being smaller than the multi-dimensional data table and having reduced dimensionality relative to the multi-dimensional data table.

2. The database management system of claim 1, wherein the multi-dimensional data table is a relational data table.

3. The database management system of claim 1, wherein the processing unit comprises:

a server application programming interface operable to provide an interface to client software;

a build unit operable to build a non-negative matrix factorization model from the data in the multi-dimensional data table; and

an apply unit operable to apply the non-negative matrix factorization model to the data in the multi-dimensional data table.

4. The database management system of claim 3, wherein the build unit comprises a table function.

5. The database management system of claim 3, wherein the build unit comprises kernel code.

6. The database management system of claim 3, wherein the apply unit comprises a table function.

7. The database management system of claim 3, wherein the apply unit comprises kernel code.

8. The database management system of claim 3, wherein the multi-dimensional data table is a relational data table.

9. The database management system of claim 3, wherein the build unit is operable to un-nest nested table columns in the multi-dimensional data table and to explode categorical attributes in the multi-dimensional data table into a series of binary attributes.

10. The database management system of claim 3, wherein the apply unit is operable to un-nest nested table columns in the multi-dimensional data table and to explode categorical attributes in the multi-dimensional data table into a series of binary attributes.

11. The database management system of claim 10, wherein the build unit is operable to un-nest nested table columns in the multi-dimensional data table and to explode categorical attributes in the multi-dimensional data table into a series of binary attributes.

12. The database management system of claim 11, wherein the multi-dimensional data table is a relational data table.

13. The database management system of claim 12, wherein the build unit comprises a table function.

14. The database management system of claim 13, wherein the apply unit comprises a table function.

15. The database management system of claim 13, wherein the apply unit comprises kernel code.

16. The database management system of claim 12, wherein the build unit comprises kernel code.

17. The database management system of claim 16, wherein the apply unit comprises a table function.

18. The database management system of claim 16, wherein the apply unit comprises kernel code.

19. A method of operating a database management system comprising:

storing data in a multi-dimensional data table; and

performing non-negative matrix factorization on data stored in the multi-dimensional data table to generate a plurality of data tables, each data table being smaller than the multi-dimensional data table and having reduced dimensionality relative to the multi-dimensional data table.

20. The method of claim 19, wherein the multi-dimensional data table is a relational data table.

21. The method of claim 19, wherein the non-negative matrix factorization includes:

providing an interface to client software;

building a non-negative matrix factorization model from the data in the multi-dimensional data table; and

applying the non-negative matrix factorization model to the data in the multi-dimensional data table.

22. The method of claim 21, wherein the non-negative matrix factorization model is built using a table function.

23. The method of claim 21, wherein the non-negative matrix factorization model is built using kernel code.

24. The method of claim 21, wherein the non-negative matrix factorization model is applied using a table function.

25. The method of claim 21, wherein the non-negative matrix factorization model is applied using kernel code.

26. The method of claim 21, wherein the multi-dimensional data table is a relational data table.

27. The method of claim 21, wherein the building step comprises:

un-nesting nested table columns in the multi-dimensional data table; and

exploding categorical attributes in the multi-dimensional data table into a series of binary attributes.

28. The method of claim 21, wherein the applying step comprises:

un-nesting nested table columns in the multi-dimensional data table; and

exploding categorical attributes in the multi-dimensional data table into a series of binary attributes.

29. The method of claim 28, wherein the building step comprises:

un-nesting nested table columns in the multi-dimensional data table; and

exploding categorical attributes in the multi-dimensional data table into a series of binary attributes.

30. The method of claim 29, wherein the multi-dimensional data table is a relational data table.

31. The method of claim 30, wherein the non-negative matrix factorization model is built using a table function.

32. The method of claim 31, wherein the non-negative matrix factorization model is applied using a table function.

33. The method of claim 31, wherein the non-negative matrix factorization model is applied using kernel code.

34. The method of claim 30, wherein the non-negative matrix factorization model is built using kernel code.

35. The method of claim 34, wherein the non-negative matrix factorization model is applied using a table function.

36. The method of claim 34, wherein the non-negative matrix factorization model is applied using kernel code.