Support vector machines in a relational database management system

• US 20050050087A1
• Filed: 08/27/2004
• Published: 03/03/2005
• 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.
• Granted Patent: US 7,565,370 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface

Abstract

An implementation of SVM functionality integrated into a relational database management system (RDBMS) improves efficiency, time consumption, and data security, reduces the parameter tuning challenges presented to the inexperienced user, and reduces the computational costs of building SVM models. A database management system comprises data stored in the database management system and a processing unit comprising a client application programming interface operable to provide an interface to client software, a build unit operable to build a support vector machine model on at least a portion of the data stored in the database management system, and an apply unit operable to apply the support vector machine model using the data stored in the database management system. The database management system may be a relational database management system.

Claims

1. A database management system comprising:

data stored in the database management system; and

a processing unit comprising;

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

a build unit operable to build a support vector machine model on at least a portion of the data stored in the database management system; and

an apply unit operable to apply the support vector machine model using the data stored in the database management system.

2. The database management system of claim 1, wherein the database management system is a relational database management system.

3. The database management system of claim 2, wherein the support vector machine model is built using a classification support vector machine algorithm.

4. The database management system of claim 2, wherein the support vector machine model is built using a regression support vector machine algorithm.

5. The database management system of claim 2, wherein the support vector machine model is built using a one-class support vector machine algorithm.

6. The database management system of claim 2, wherein the build unit is further operable to build a support vector machine model on at least a portion of the data stored in the database management system using data sampling, and the apply unit is further operable to apply the support vector machine model using the data stored in the database management system using models based on data sampling.

7. The database management system of claim 6, wherein the data sampling may comprise simple or stratified random sampling, or active sampling.

8. The database management system of claim 6, wherein the support vector machine model is built using a classification support vector machine algorithm, a regression support vector machine algorithm, or a one-class support vector machine algorithm.

9. The database management system of claim 2, wherein the build unit is further operable to build a support vector machine model on at least a portion of the data stored in the database management system using a data summarization driven method, and the apply unit is further operable to apply the support vector machine model using the data stored in the database management system using models based on data summarization driven method.

10. The database management system of claim 2, wherein the build unit is further operable to build a support vector machine model on at least a portion of the data stored in the database management system using stratified random sampling, and the apply unit is further operable to apply the support vector machine model using the data stored in the database management system using stratified random sampling.

11. The database management system of claim 2, wherein the processing unit further comprises an incremental update unit.

12. The database management system of claim 11, wherein the incremental update unit is operable to incorporate additional data into an existing model without rebuilding the entire model.

13. The database management system of claim 11, wherein the incremental update unit is operable to adapt an existing model to a stream of data wherein an underlying distribution changes over time.

14. The database management system of claim 1, wherein the build unit comprises a table function

15. The database management system of claim 1, wherein the build unit comprises kernel code.

16. The database management system of claim 1, wherein the apply unit comprises a table function.

17. The database management system of claim 1, wherein the apply unit comprises kernel code.

18. The database management system of claim 2, wherein the build unit is further operable to perform partial random chunking selection by:

i) selecting an initial chunk of data from the data stored in the database management system;

ii) performing optimization on the selected chunk until the optimization converges on the selected chunk according to convergence conditions;

iii) checking the convergence conditions against the data stored in the database management system to obtain violators of the convergence conditions;

iv) updating the chunk to form a new selected chunk comprising members of the previous selected chunk and at least some of violators of the convergence conditions, selected randomly with a bias towards large violators; and

v) repeating ii-iv until the convergence conditions are met on at least a portion of the data stored in the database management system.

19. The database management system of claim 18, wherein the support vector machine model is built using a classification support vector machine algorithm, a regression support vector machine algorithm, or a one-class support vector machine algorithm.

20. The database management system of claim 18, wherein the build unit is further operable to build a support vector machine model on at least a portion of the data stored in the database management system using data sampling, and the apply unit is further operable to apply the support vector machine model using the data stored in the database management system using models based on data sampling.

21. The database management system of claim 20, wherein the build unit uses an active sampling method.

22. The database management system of claim 20, wherein the support vector machine model is built using a classification support vector machine algorithm, a regression support vector machine algorithm, or a one-class support vector machine algorithm.

23. The database management system of claim 18, wherein the build unit is further operable to build a support vector machine model on at least a portion of the data stored in the database management system using a data summarization driven method, and the apply unit is further operable to apply the support vector machine model using the data stored in the database management system using models based on data summarization driven method.

24. The database management system of claim 18, wherein the build unit is further operable to build a support vector machine model on at least a portion of the data stored in the database management system using stratified random sampling, and the apply unit is further operable to apply the support vector machine model using the data stored in the database management system using stratified random sampling.

25. The database management system of claim 18, wherein the processing unit further comprises an incremental update unit.

26. The database management system of claim 25, wherein the incremental update unit is operable to incorporate additional data into an existing model without rebuilding the entire model.

27. The database management system of claim 25, wherein the incremental update unit is operable to adapt an existing model to a stream of data wherein an underlying distribution changes over time.

28. The database management system of claim 2, wherein the build unit is further operable to perform partial random chunking selection by:

i) selecting a pool of data from the data stored in the database management system;

ii) selecting an initial chunk of data from the pool of data;

iii) performing optimization on the selected chunk until the optimization converges on the selected chunk according to convergence conditions;

iv) checking the convergence conditions against the pool of data to obtain violators of the convergence conditions;

v) updating the chunk to form a new selected chunk comprising members of the previous selected chunk and a smallest violator of the convergence conditions, if the size of the chunk does not exceed a threshold; and

vi) repeating iii-v until the convergence conditions are met on at least a portion of the data pool.

29. The database management system of claim 2, wherein the support vector machine model is built using a classification support vector machine algorithm, a regression support vector machine algorithm, or a one-class support vector machine algorithm.

30. The database management system of claim 28, wherein the build unit is further operable to build a support vector machine model on at least a portion of the data stored in the database management system using data sampling, and the apply unit is further operable to apply the support vector machine model using the data stored in the database management system using models based on data sampling.

31. The database management system of claim 30, wherein the build unit uses an active sampling method.

32. The database management system of claim 30, wherein the support vector machine model is built using a classification support vector machine algorithm, a regression support vector machine algorithm, or a one-class support vector machine algorithm.

33. The database management system of claim 28, wherein the build unit is further operable to build a support vector machine model on at least a portion of the data stored in the database management system using a data summarization driven method, and the apply unit is further operable to apply the support vector machine model using the data stored in the database management system using models based on data summarization driven method.

34. The database management system of claim 28, wherein the build unit is further operable to build a support vector machine model on at least a portion of the data stored in the database management system using stratified random sampling, and the apply unit is further operable to apply the support vector machine model using the data stored in the database management system using stratified random sampling.

35. The database management system of claim 28, wherein the processing unit further comprises an incremental update unit.

36. The database management system of claim 35, wherein the incremental update unit is operable to incorporate additional data into an existing model without rebuilding the entire model.

37. The database management system of claim 35, wherein the incremental update unit is operable to adapt an existing model to a stream of data wherein an underlying distribution changes over time.

39. The method of claim 38, wherein the database management system is a relational database management system.

40. The method of claim 39, wherein the support vector machine model is built using a classification support vector machine algorithm.

41. The method of claim 39, wherein the support vector machine model is built using a regression support vector machine algorithm.

42. The method of claim 39, wherein the building step comprises using a one-class support vector machine algorithm.

43. The method of claim 39, wherein the building step comprises building a support vector machine model on at least a portion of the data stored in the database management system using data sampling, and the applying step comprises applying the support vector machine model using the data stored in the database management system using models based on data sampling.

44. The method of claim 43, wherein the data sampling comprises simple or stratified random sampling, or active sampling.

45. The method of claim 43, wherein the building step comprises using a classification support vector machine algorithm, a regression support vector machine algorithm, or a one-class support vector machine algorithm.

46. The method of claim 39, wherein the building step comprises building a support vector machine model on at least a portion of the data stored in the database management system using a data summarization driven method, and applying step comprises applying the support vector machine model using the data stored in the database management system using models based on data summarization driven method.

47. The method of claim 39, wherein the building step comprises building a support vector machine model on at least a portion of the data stored in the database management system using stratified random sampling, and applying step comprises applying the support vector machine model using the data stored in the database management system using stratified random sampling.

48. The method of claim 39, wherein the building step further comprises incrementally updating an existing model to incorporate additional data into the model without rebuilding the entire model.

49. The method of claim 48, wherein the incrementally updating step comprises adapting an existing model to a stream of data wherein an underlying distribution changes over time.

54. The method of claim 39, wherein the build step comprises performing partial random chunking selection by:

i) selecting an initial chunk of data from the data stored in the database management system;

ii) performing optimization on the selected chunk until the optimization converges on the selected chunk according to convergence conditions;

iii) checking the convergence conditions against the data stored in the database management system to obtain violators of the convergence conditions;

iv) updating the chunk to form a new selected chunk comprising members of the previous selected chunk and at least some of violators of the convergence conditions, selected randomly with a bias towards large violators; and

v) repeating ii-iv until the convergence conditions are met on at least a portion of the data stored in the database management system.

55. The method of claim 54, wherein the building step comprises using a classification support vector machine algorithm, a regression support vector machine algorithm, or a one-class support vector machine algorithm.

56. The method of claim 54, wherein the building step comprises building a support vector machine model on at least a portion of the data stored in the database management system using data sampling, and the applying step comprises applying the support vector machine model using the data stored in the database management system using models based on data sampling.

57. The method of claim 56, wherein the building step comprises using an active sampling method.

58. The method of claim 56, wherein the building step comprises using a classification support vector machine algorithm, a regression support vector machine algorithm, or a one-class support vector machine algorithm.

59. The method of claim 54, wherein the building step comprises building a support vector machine model on at least a portion of the data stored in the database management system using a data summarization driven method, and applying step comprises applying the support vector machine model using the data stored in the database management system using models based on data summarization driven method.

60. The method of claim 54, wherein the building step comprises building a support vector machine model on at least a portion of the data stored in the database management system using stratified random sampling, and applying step comprises applying the support vector machine model using the data stored in the database management system using stratified random sampling.

61. The method of claim 54, wherein the building step further comprises incrementally updating an existing model to incorporate additional data into the model without rebuilding the entire model.

62. The method of claim 61, wherein the incrementally updating step comprises adapting an existing model to a stream of data wherein an underlying distribution changes over time.

63. The method of claim 39, wherein the build step comprises performing partial random chunking selection by:

i) selecting a pool of data from the data stored in the database management system;

ii) selecting an initial chunk of data from the pool of data;

iii) performing optimization on the selected chunk until the optimization converges on the selected chunk according to convergence conditions;

iv) checking the convergence conditions against the pool of data to obtain violators of the convergence conditions;

v) updating the chunk to form a new selected chunk comprising members of the previous selected chunk and a smallest violator of the convergence conditions, if the size of the chunk does not exceed a threshold; and

vi) repeating iii-v until the convergence conditions are met on at least a portion of the data pool.

64. The method of claim 63, wherein the building step comprises using a classification support vector machine algorithm, a regression support vector machine algorithm, or a one-class support vector machine algorithm.

65. The method of claim 63, wherein the building step comprises building a support vector machine model on at least a portion of the data stored in the database management system using data sampling, and the applying step comprises applying the support vector machine model using the data stored in the database management system using models based on data sampling.

66. The method of claim 65, wherein the building step comprises using an active sampling method.

67. The method of claim 65, wherein the building step comprises using a classification support vector machine algorithm, a regression support vector machine algorithm, or a one-class support vector machine algorithm.

68. The method of claim 63, wherein the building step comprises building a support vector machine model on at least a portion of the data stored in the database management system using a data summarization driven method, and applying step comprises applying the support vector machine model using the data stored in the database management system using models based on data summarization driven method.

69. The method of claim 63, wherein the building step comprises building a support vector machine model on at least a portion of the data stored in the database management system using stratified random sampling, and applying step comprises applying the support vector machine model using the data stored in the database management system using stratified random sampling.

70. The method of claim 63, wherein the building step further comprises incrementally updating an existing model to incorporate additional data into the model without rebuilding the entire model.

71. The method of claim 70, wherein the incrementally updating step comprises adapting an existing model to a stream of data wherein an underlying distribution changes over time.

38. In a database management system, a method comprising:

storing data in the database management system;

providing an interface to client software;

building a support vector machine model on at least a portion of the data stored in the database management system; and

applying the support vector machine model using the data stored in the database management system.

50. The method of claim 38, wherein the building step comprises using a table function

51. The method of claim 38, wherein the building step comprises using kernel code.

52. The method of claim 38, wherein the apply step comprises using a table function.

53. The method of claim 38, wherein the apply step comprises using kernel code.