Methods and apparatus to utilize minimum cross entropy to calculate granular data of a region based on another region for media audience measurement
View Patent ↗Methods and apparatus for calculating granular data of an audience of a target region. An example method includes utilizing minimum cross entropy to apportion aggregate demographics data and aggregate behavioral data of the target region based on granular data of a source region. The example further includes reducing an amount of computer memory resources by calculating the granular data of the audience without collecting person-specific data from audience members of the target region, the granular data of the audience including (A) a count of the audience of the target region satisfying a behavioral constraint and a first demographic constraint and (B) a count of the audience of the target region satisfying the behavioral constraint and a second demographic constraint, the calculated granular data based on the aggregate demographics data of the target region, the aggregate behavioral data of the target region, and the granular data of the source region.
1. An apparatus for calculating granular data of an audience of a target region, the apparatus comprising:
a means for utilizing minimum cross entropy to apportion aggregate demographics data and aggregate behavioral data of the target region based on granular data of a source region; and
a means for reducing an amount of computer memory resources by calculating the granular data of the audience of the target region without collecting person-specific data from audience members in the audience of the target region, the granular data of the audience of the target region including (A) a count of a first subset of the audience of the target region satisfying a behavioral constraint and a first demographic constraint and (B) a count of a second subset of the audience of the target region satisfying the behavioral constraint and a second demographic constraint, the calculated granular data of the audience of the target region based on the apportioned aggregate demographics data of the target region, the apportioned aggregate behavioral data of the target region, and the granular data of the source region.
2. The apparatus as defined in claim 1 , wherein the apportioned aggregate behavioral data of the target region includes tuning data of the target region and the granular data of the source region includes exposure data of the source region, the target region being a non-panelist region and the source region being a panelist region, the non-panelist region and the panelist region being mutually exclusive.
3. The apparatus as defined in claim 1 , wherein the aggregate behavioral data of the target region is non-person-specific data of a non-panelist region that indicates a quantity of members of the non-panelist region that satisfy the behavioral constraint, and the aggregate demographic data of the target region is non-person-specific data of the non-panelist region that indicates a quantity of members of the non-panelist region that satisfy the demographic constraints.
4. The apparatus as defined in claim 1 , wherein the calculation of the granular data of the audience of the target region further includes means for performing non-linear optimization based on at least one of (A) shared behavioral characteristics in the granular data of the source region and in the aggregate behavioral data of the target region, or (B) shared demographic characteristics in the granular data of the source region and in the aggregate demographics data of the target region.
5. The apparatus as defined in claim 1 , wherein the calculation of the granular data of the audience of the target region further includes:
means for performing non-linear optimization based on the granular data of the source region, the aggregate demographics data of the target region, and the aggregate behavioral data of the target region; and
means for defining an optimization constraint based on the aggregate behavioral data and the aggregate demographics data of the target region, the non-linear optimization being limited by the optimization constraint.
6. The apparatus as defined in claim 1 , further including means for increasing accuracy of values of the granular data of the audience of the target region by applying the minimum cross entropy to reduce variability of the values of the audience of the target region, the increasing accuracy means to cause greater differences in intermediate values as compared to values of the granular data of the audience of the target region proximate to at least one of 0% or 100%, the intermediate values proximate to 50%.
7. The apparatus as defined in claim 1 , wherein the first demographic constraint and the second demographic constraint are mutually exclusive.
8. A computer implemented method to calculate granular data of an audience of a target region, the computer implemented method comprising:
utilizing, by executing an instruction with a processor, minimum cross entropy to apportion aggregate demographics data and aggregate behavioral data of the target region based on granular data of a source region; and
reducing, by executing an instruction with the processor, an amount of computer memory resources by calculating the granular data of the audience of the target region without collecting person-specific data from audience members in the audience of the target region, the granular data of the audience of the target region including (A) a count of a first subset of the audience of the target region satisfying a behavioral constraint and a first demographic constraint and (B) a count of a second subset of the audience of the target region satisfying the behavioral constraint and a second demographic constraint, the calculated granular data of the audience of the target region based on the apportioned aggregate demographics data of the target region, the apportioned aggregate behavioral data of the target region, and the granular data of the source region.
9. The method as defined in claim 8 , wherein the apportioned aggregate behavioral data of the target region includes tuning data of the target region, and the granular data of the source region includes exposure data of the source region, the target region being a non-panelist region and the source region being a panelist region, the non-panelist region and the panelist region being mutually exclusive.
10. The method as defined in claim 8 , wherein the aggregate behavioral data is non-person-specific data of a non-panelist region that indicates a quantity of members of the non-panelist region that satisfy the behavioral constraint, and the aggregate demographic data is non-person-specific data of the non-panelist region that indicates a quantity of members of the non-panelist region that satisfy the demographic constraints.
11. The method as defined in claim 8 , wherein the calculation of the granular data of the audience of the target region further includes performing non-linear optimization based on at least one of (A) shared behavioral characteristics in the granular data of the source region and in the aggregate behavioral data of the target region, or (B) shared demographic characteristics in the granular data of the source region and in the aggregate demographics data of the target region.
12. The method as defined in claim 8 , wherein the calculation of the granular data of the audience of the target region further includes:
performing non-linear optimization based on the granular data of the source region, the aggregate demographics data of the target region, and the aggregate behavioral data of the target region; and
defining an optimization constraint based on the aggregate behavioral data and the aggregate demographics data of the target region, the non-linear optimization being limited by the optimization constraint.
13. The method as defined in claim 8 , wherein the accuracy of values of the granular data of the audience of the target region are increased by applying the minimum cross entropy to reduce variability of the values of the audience of the target region and to cause greater differences in intermediate values as compared to values of the granular data of the audience of the target region proximate to at least one of 0% or 100%, the intermediate values proximate to 50%.
14. The method as defined in claim 8 , wherein the first demographic constraint and the second demographic constraint are mutually exclusive.
15. A tangible computer readable storage medium to calculate granular data of an audience of a target region, the tangible computer readable storage medium comprising instructions which, when executed, cause a machine to at least:
utilize minimum cross entropy to apportion aggregate demographics data and aggregate behavioral data of the target region based on granular data of a source region; and
reduce an amount of computer memory resources by calculating the granular data of the audience of the target region without collecting person-specific data from audience members in the audience of the target region, the granular data of the audience of the target region including (A) a count of a first subset of the audience of the target region satisfying a behavioral constraint and a first demographic constraint and (B) a count of a second subset of the audience of the target region satisfying the behavioral constraint and a second demographic constraint, the calculated granular data of the audience of the target region based on the apportioned aggregate demographics data of the target region, the apportioned aggregate behavioral data of the target region, and the granular data of the source region.
16. The tangible computer readable storage medium as defined in claim 15 , wherein the apportioned aggregate behavioral data of the target region includes tuning data of the target region and the granular data of the source region includes exposure data of the source region, the target region being a non-panelist region and the source region being a panelist region, the non-panelist region and the panelist region being mutually exclusive.
17. The tangible computer readable storage medium as defined in claim 15 , wherein the aggregate behavioral data is non-person-specific data of a non-panelist region that indicates a quantity of members of the non-panelist region that satisfy the behavioral constraint, and the aggregate demographic data is non-person-specific data of the non-panelist region that indicates a quantity of members of the non-panelist region that satisfy the demographic constraints.
18. The tangible computer readable storage medium as defined in claim 15 , wherein the calculation of the granular data of the audience of the target region further includes performing non-linear optimization based on at least one of (A) shared behavioral characteristics in the granular data of the source region and in the aggregate behavioral data of the target region, or (B) shared demographic characteristics in the granular data of the source region and in the aggregate demographics data of the target region.
19. The tangible computer readable storage medium as defined in claim 15 , wherein the instructions, when executed, further cause the machine to:
calculate the granular data of the audience of the target region by:
performing non-linear optimization based on the granular data of the source region, the aggregate demographics data of the target region, and the aggregate behavioral data of the target region; and
defining an optimization constraint based on the aggregate behavioral data and the aggregate demographics data of the target region, the non-linear optimization being limited by the optimization constraint.
20. The tangible computer readable storage medium as defined in claim 15 , wherein the accuracy of values of the granular data of the audience of the target region are increased by applying the minimum cross entropy to reduce variability of the values of the audience of the target region and to cause greater differences in intermediate values as compared to values of the granular data of the audience of the target region proximate to at least one of 0% or 100%, the intermediate values proximate to 50%.