Methods, systems and apparatus to improve multi-demographic modeling efficiency

1. An apparatus to reduce computational resources for a multi-demographic training model, comprising:

a feature set aggregator to:

segregate training data based on feature sets of interest; and

identify households that participate in at least one of the feature sets of interest;

a class enumerator to reduce multi-demographic model iterations by enumerating demographic combinations for the identified households, the enumerated demographic combinations including a single identifier to represent a combination of two or more demographic categories;

a linear combiner to:

select a demographic category of interest from the two or more demographic categories;

discriminate which ones of the enumerated demographic combinations are to be selected based on whether the selected demographic category of interest is present therein; and

when the selected demographic category of interest is present, select the respective ones of the enumerated demographic combinations based on the single identifier, the single identifier including cartesian representations that reflect occurrences or non-occurrences of the selected demographic category of interest; and

a modeling engine to generate training coefficients corresponding to the selected ones of the enumerated demographic combinations.

2. The apparatus as defined in claim 1 , further including a household data associator to associate demographic combinations with respective ones of the identified households.

3. The apparatus as defined in claim 1 , wherein the modeling engine is to perform a multinomial logistic regression with the segregated training data and the selected ones of the enumerated demographic combinations to generate the training coefficients.

4. The apparatus as defined in claim 1 , wherein the modeling engine is to generate probability values for the selected ones of the enumerated demographic combinations.

5. The apparatus as defined in claim 4 , wherein the linear combiner is to revert the probability values for respective ones of the enumerated demographic combinations to probability values associated with individual demographic components.

6. The apparatus as defined in claim 1 , further including a data retriever to retrieve a third party model, the modeling engine to generate probability values based on the segregated training data for the selected ones of the enumerated demographic combinations associated with (a) a training model and (b) the third party model.

7. The apparatus as defined in claim 6 , wherein the linear combiner is to determine probability value differences between the training model and the third party model.

8. The apparatus as defined in claim 7 , wherein the linear combiner is to compare the probability value differences to a threshold to determine a trust metric for the third party model.

9. A computer-implemented method to reduce computational resources for a multi-demographic training model, the method comprising:

segregating, by executing an instruction with a processor, training data based on feature sets of interest;

identifying, by executing an instruction with the processor, households that participate in at least one of the feature sets of interest;

reducing, by executing an instruction with the processor, multi-demographic model iterations by enumerating demographic combinations for the identified households, the enumerated demographic combinations including a single identifier to represent a combination of two or more demographic categories;

selecting a demographic category of interest from the two or more demographic categories;

discriminating which ones of the enumerated demographic combinations to select based on whether the selected demographic category of interest is present therein;

when the selected demographic category of interest is present, selecting respective ones of the enumerated demographic combinations based on the single identifier, the single identifier including cartesian representations that reflect occurrences or non-occurrences of the selected demographic category of interest; and

generating, by executing an instruction with the processor, training coefficients corresponding to the selected ones of the enumerated demographic combinations.

10. A method as defined in claim 9 , further including associating demographic combinations with respective ones of the identified households.

11. A method as defined in claim 9 , further including performing a multinomial logistic regression with the segregated training data and the selected ones of the enumerated demographic combinations to generate the training coefficients.

12. A method as defined in claim 9 , further including generating probability values for the selected ones of the enumerated demographic combinations.

13. A method as defined in claim 12 , further including reverting the probability values for respective ones of the enumerated demographic combinations to probability values associated with individual demographic components.

14. A method as defined in claim 9 , further including retrieving a third party model, and wherein the generating probability values is based on the segregated training data for the selected ones of the enumerated demographic combinations associated with (a) a training model and (b) the third party model.

15. A method as defined in claim 14 , further including determining probability value differences between the training model and the third party model.

16. A method as defined in claim 15 , further including comparing the probability value differences to a threshold to determine a trust metric for the third party model.

17. A tangible computer-readable storage medium comprising instructions that, when executed, cause a processor to, at least:

segregate training data based on feature sets of interest;

identify households that participate in at least one of the feature sets of interest;

reduce multi-demographic model iterations by enumerating demographic combinations for the identified households, the enumerated demographic combinations including a single identifier to represent a combination of two or more demographic categories;

select a demographic category of interest from the two or more demographic categories;

discriminate which ones of the enumerated demographic combinations to select based on whether the selected demographic category of interest is present therein;

when the selected demographic category of interest is present, select respective ones of the enumerated demographic combination based on the single identifier, the single identifier including cartesian representations that reflect occurrences or non-occurrences of the selected demographic category of interest; and

generate training coefficients corresponding to the selected ones of the enumerated demographic combinations.

18. The computer-readable medium as defined in claim 17 including instructions that, when executed, cause the processor to associate demographic combinations with respective ones of the identified households.

19. The computer-readable medium as defined in claim 17 including instructions that, when executed, cause the processor to perform a multinomial logistic regression with the segregated training data and the selected ones of the enumerated demographic combinations to generate the training coefficients.

20. The computer-readable medium as defined in claim 17 including instructions that, when executed, cause the processor to generate probability values for the selected ones of the enumerated demographic combinations.