Approximating value densities

• US 9,471,545 B2
• Filed: 02/11/2013
• Issued: 10/18/2016
• Est. Priority Date: 02/11/2013
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Milenova, Boriana Lubomirova, Campos, Marcos M, Yarmus, Joseph
• Primary Examiner(s): Alhija, Saif
• Application Number:
  US13/764,677
• Publication Number:
  US 20140229147A1
• Time in Patent Office: 1,345 Days
• Field of Search: 703/2
• US Class Current: 1/1
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04

Abstract

Processes, machines, and stored machine instructions are provided for approximating value densities in data. While generating a resulting density model to approximate value densities in a set of data, density modeling logic selects a functional component of a first model to vary based at least in part on how much the functional component contributes to how well the first model approximates the value densities. The density modeling logic then uses at least the functional component and a variation of the functional component as seed components to determine adjusted functional components of a second model by iteratively determining, in an expectation step, how much the seed components contribute to how well the second model explains the values, and, in a maximization step, new seed components, optionally to be used in further iterations, based at least in part on how much of the values are attributable to the seed components.

Claims

1. A method comprising:

generating a first density model that approximates value densities in a set of data, wherein the first density model comprises a first set of functional components;

iteratively generating at least one additional density model comprising a second density model, wherein generating the second density model comprises;

selecting a first functional component of the first set of functional components based at least in part on how much the first functional component contributes to how well the first set of functional components approximates the value densities in the set of data,

generating a variation of the first functional component, and

generating the second density model comprising a second set of functional components, wherein the second set of functional components includes at least one more functional component than the first set of functional components, and wherein the second set of functional components is determined using at least the first functional component and the variation of the first functional component as seed components;

storing, in association with the set of data, a resulting density model selected from the at least one additional density model;

wherein the method is performed by one or more computing devices.

2. The method of claim 1, wherein iteratively generating the at least one additional density model further comprises:

determining an amount of improvement that indicates how much better the second density model approximates the value densities in the set of data than the first density model approximates the value densities in the set of data;

based at least in part on determining that the amount of improvement does not satisfy a threshold amount of improvement;

selecting a second functional component of the first set of functional components, wherein the second functional component is different than the first functional component; and

generating a third density model comprising a third set of functional components, wherein the third set of functional components includes at least one more functional component than the first set of functional components, and wherein the third set of functional components is determined using at least the second functional component and a variation of the second functional component as seed components;

wherein the at least one additional density model comprises the third density model.

3. The method of claim 1, wherein iteratively generating the at least one additional density model further comprises:

determining an amount of improvement that indicates how much better the second density model approximates the value densities in the set of data than the first density model approximates the value densities in the set of data;

based at least in part on determining that the amount of improvement does not satisfy a threshold amount of improvement, stop iteratively generating the at least one additional density model.

4. The method of claim 1, wherein iteratively generating the at least one additional density model further comprises:

determining an amount of improvement that indicates how much better the second density model approximates the value densities in the set of data than the first density model approximates the value densities in the set of data;

based at least in part on determining that the amount of improvement satisfies a threshold amount of improvement;

selecting a second functional component of the second set of functional components based at least in part on how much the second functional component contributes to how well the second set of functional components approximates the value densities in the set of data; and

generating a third density model comprising a third set of functional components, wherein the third set of functional components includes at least one more functional component than the second set of functional components, and wherein the third set of functional components is determined using at least the second functional component and a variation of the second functional component as seed components;

wherein the at least one additional density model comprises the third density model.

5. The method of claim 1, wherein iteratively generating the at least one additional density model further comprises:

determining whether a number of components in the second density model is a threshold number of components allowed for the resulting density model;

based at least in part on determining that the number of components in the second density model is less than the threshold number of components;

selecting a second functional component of the second set of functional components based at least in part on how much the second functional component contributes to how well the second set of functional components approximates the value densities in the set of data; and

generating a third density model comprising a third set of functional components, wherein the third set of functional components includes at least one more functional component than the first set of functional components, and wherein the third set of functional components is determined using at least the second functional component and a variation of the second functional component as seed components;

wherein the at least one additional density model comprises the third density model.

6. The method of claim 1, wherein iteratively generating the at least one additional density model further comprises:

determining whether a number of components in the second density model is a threshold number of components allowed for the resulting density model;

based at least in part on determining that the number of components in the second density model meets the threshold number of components, stop iteratively generating the at least one additional density model.

7. The method of claim 1, further comprising:

determining a degree of accuracy or relative degree of improvement by which the second density model approximates the value densities in the set of data;

wherein storing the resulting density model comprises storing the second density model based at least in part on determining that the degree of accuracy or relative degree of improvement satisfies a threshold.

8. The method of claim 1, wherein the seed components also include one or more other components in a neighborhood of the first functional component, and wherein determining the second set of functional components comprises determining adjusted functional components of the second density model by determining, in an expectation step, how much the seed components contribute to how well the second density model approximates value distributions, and, in a maximization step, updated seed components based at least in part on how much of at least part of the set of data is attributable to the seed components.

9. The method of claim 1, wherein determining the second set of functional components using at least the first functional component and the variation of the first functional component as seed components comprises determining adjusted functional components of the second density model by determining, in an expectation step, how much the seed components contribute to how well the second density model approximates value distributions, and, in a maximization step, updated seed components based at least in part on how much of at least part of the set of data is attributable to the seed components.

10. The method of claim 1, wherein the set of data comprises a binned representation of a raw set of data that comprises non-numerical values.

11. The method of claim 1, wherein the first functional component is a Gaussian distribution centered at a specified value, and wherein the variation of the first functional component is generated by varying a location parameter of the first functional component in a direction of maximum variance.

12. The method of claim 1, wherein the first functional component is a multinomial distribution comprising one or more frequencies, and wherein the variation of the first functional component is generated by varying the one or more frequencies of the multinomial distribution.

13. The method of claim 1, wherein how well the first set of functional components approximates the value densities in the set of data is measured based at least in part on a likelihood that at least part of the set of data is explained by the first set of functional components.

14. One or more non-transitory storage media storing instructions which, when executed by one or more computing devices, cause:

generating a first density model that approximates value densities in a set of data, wherein the first density model comprises a first set of functional components;

iteratively generating at least one additional density model comprising a second density model, wherein generating the second density model comprises;

selecting a first functional component of the first set of functional components based at least in part on how much the first functional component contributes to how well the first set of functional components approximates the value densities in the set of data,

generating a variation of the first functional component, and

generating a second density model comprising a second set of functional components, wherein the second set of functional components includes at least one more functional component than the first set of functional components, and wherein the second set of functional components is determined using at least the first functional component and the variation of the first functional component as seed components;

storing, in association with the set of data, a resulting density model selected from the at least one additional density model.

15. The one or more non-transitory storage media of claim 14, wherein iteratively generating the at least one additional density model further comprises:

determining an amount of improvement that indicates how much better the second density model approximates the value densities in the set of data than the first density model approximates the value densities in the set of data;

based at least in part on determining that the amount of improvement does not satisfy a threshold amount of improvement;

selecting a second functional component of the first set of functional components, wherein the second functional component is different than the first functional component; and

generating a third density model comprising a third set of functional components, wherein the third set of functional components includes at least one more functional component than the first set of functional components, and wherein the third set of functional components is determined using at least the second functional component and a variation of the second functional component as seed components;

wherein the at least one additional density model comprises the third density model.

16. The one or more non-transitory storage media of claim 14, wherein iteratively generating the at least one additional density model further comprises:

determining an amount of improvement that indicates how much better the second density model approximates the value densities in the set of data than the first density model approximates the value densities in the set of data;

based at least in part on determining that the amount of improvement does not satisfy a threshold amount of improvement, stop iteratively generating the at least one additional density model.

17. The one or more non-transitory storage media of claim 14, wherein iteratively generating the at least one additional density model further comprises:

determining an amount of improvement that indicates how much better the second density model approximates the value densities in the set of data than the first density model approximates the value densities in the set of data;

based at least in part on determining that the amount of improvement satisfies a threshold amount of improvement;

selecting a second functional component of the second set of functional components based at least in part on how much the second functional component contributes to how well the second set of functional components approximates the value densities in the set of data; and

generating a third density model comprising a third set of functional components, wherein the third set of functional components includes at least one more functional component than the second set of functional components, and wherein the third set of functional components is determined using at least the second functional component and a variation of the second functional component as seed components;

wherein the at least one additional density model comprises the third density model.

18. The one or more non-transitory storage media of claim 14, wherein iteratively generating the at least one additional density model further comprises:

determining whether a number of components in the second density model is a threshold number of components allowed for the resulting density model;

based at least in part on determining that the number of components in the second density model is less than the threshold number of components;

selecting a second functional component of the second set of functional components based at least in part on how much the second functional component contributes to how well the second set of functional components approximates the value densities in the set of data; and

generating a third density model comprising a third set of functional components, wherein the third set of functional components includes at least one more functional component than the first set of functional components, and wherein the third set of functional components is determined using at least the second functional component and a variation of the second functional component as seed components;

wherein the at least one additional density model comprises the third density model.

19. The one or more non-transitory storage media of claim 14, wherein iteratively generating the at least one additional density model further comprises:

determining whether a number of components in the second density model is a threshold number of components allowed for the resulting density model;

based at least in part on determining that the number of components in the second density model meets the threshold number of components, stop iteratively generating the at least one additional density model.

20. The one or more non-transitory storage media of claim 14, wherein the instructions, when executed, further cause:

determining a degree of accuracy or relative degree of improvement by which the second density model approximates the value densities in the set of data;

wherein storing the resulting density model comprises storing the second density model based at least in part on determining that the degree of accuracy or relative degree of improvement satisfies a threshold.

21. The one or more non-transitory storage media of claim 14, wherein the seed components also include one or more other components in a neighborhood of the first functional component, and wherein determining the second set of functional components comprises determining adjusted functional components of the second density model by determining, in an expectation step, how much the seed components contribute to how well the second density model approximates value distributions, and, in a maximization step, updated seed components based at least in part on how much of at least part of the set of data is attributable to the seed components.

22. The one or more non-transitory storage media of claim 14, wherein determining the second set of functional components using at least the first functional component and the variation of the first functional component as seed components comprises determining adjusted functional components of the second density model by determining, in an expectation step, how much the seed components contribute to how well the second density model approximates value distributions, and, in a maximization step, updated seed components based at least in part on how much of at least part of the set of data is attributable to the seed components.

23. The one or more non-transitory storage media of claim 14, wherein the set of data comprises a binned representation of a raw set of data that comprises non-numerical values.

24. The one or more non-transitory storage media of claim 14, wherein the first functional component is a Gaussian distribution centered at a specified value, and wherein the variation of the first functional component is generated by varying a location parameter of the first functional component in a direction of maximum variance.

25. The one or more non-transitory storage media of claim 14, wherein the first functional component is a multinomial distribution comprising one or more frequencies, and wherein the variation of the first functional component is generated by varying the one or more frequencies of the multinomial distribution.

26. The one or more non-transitory storage media of claim 14, wherein how well the first set of functional components approximates the value densities in the set of data is measured based at least in part on a likelihood that at least part of the set of data is explained by the first set of functional components.