Synthetic total audience ratings

1 . A method, comprising:

accessing census data including census viewing statements associated with media content presented by a plurality of census devices;

accessing panel data including digital viewing activity and panelist viewing statements associated with media content presented by a plurality of panel devices, the panel data further including panel demographic information associated with each panel device of the plurality of panel devices, wherein accessing the panel data comprises:

accessing the panel data from at least one streaming meter configured to measure the digital viewing activity via a broadband connection on each panel device of the plurality of panel devices in a respective broadband capable panel household, wherein the at least one streaming meter is connected to a household WiFi network in the respective broadband capable panel household, wherein each panel device of the plurality of panel devices in the respective broadband capable panel household is registered with a primary device user and mapped to the at least one streaming meter, the digital viewing activity comprising digital event information including the primary device user for each panel device of the plurality of panel devices;

determining scores representing similarities between ones of a first group of census devices determined to have matching panel devices in the panel data and ones of a second group of census devices determined to be unmatched in the panel data; and

assigning each census device of the plurality of census devices a respective demographic class based on the panel demographic information and the scores representing the similarities.

2 . The method of claim 1 , wherein determining the scores representing the similarities between the ones of the first group of census devices determined to have the matching panel devices in the panel data and the ones of the second group of census devices determined to be unmatched in the panel data comprises:

implementing a demographic attribution process based on at least one of a recurrent neural network, a Siamese network, or geometric distance calculations to determine the scores representing the similarities.

3 . The method of claim 1 , wherein the respective demographic class comprises at least one of age, gender, education, working status, city size, or household size.

4 . The method of claim 1 , wherein assigning each census device of the plurality of census devices the respective demographic class comprises applying a machine learning (ML) model trained on the panel demographic information to the census devices.

5 . The method of claim 1 , wherein respective numbers of census devices assigned to corresponding ones of the panel devices is in accordance with weights of the corresponding ones of the panel devices.

6 . A method comprising:

accessing census data including census viewing statements associated with media content presented by a plurality of census devices;

accessing panel data including panelist viewing statements associated with media content presented by a plurality of panel devices, the panel data further including panel demographic information associated with each panel device of the plurality of panel devices;

determining scores representing similarities between ones of a first group of census devices determined to have matching panel devices in the panel data and ones of a second group of census devices determined to be unmatched in the panel data; and

assigning each census device of the plurality of census devices a respective demographic class based on the panel demographic information and the scores representing the similarities, wherein respective numbers of census devices assigned to corresponding ones of the panel devices is in accordance with weights of the corresponding ones of the panel devices, wherein the panel data including the weights that are representative of numbers of individuals in a population to be represented by corresponding panelists represented in the panel data, the method further comprising:

dividing the weights for respective ones of the panelists into respective sets of split weights; and

assigning census viewing statements to at least subsets of the sets of split weights to determine the audience ratings for the population, the assignment of the census viewing statements to the at least subsets of the sets of split weights based on a reach constraint.

7 . A computing system comprising:

a processor;

at least one streaming meter configured to measure digital viewing activity via a broadband connection presented on the plurality of panel devices in a respective broadband capable panel household, wherein the at least one streaming meter is connected to a household WiFi network in the respective broadband capable panel household, wherein each panel device of the plurality of panel devices in the respective broadband capable panel household is registered with a primary device user and mapped to the at least one streaming meter, the digital viewing activity comprising digital event information including the primary device user for each panel device of the plurality of panel devices; and

a non-transitory computer-readable storage medium, having stored thereon program instructions that, upon execution by the processor, cause performance of a set of operations comprising:

accessing census data including census viewing statements associated with media content presented by a plurality of census devices;

accessing panel data including the digital event information and panelist viewing statements associated with media content presented by the plurality of panel devices, the panel data further including panel demographic information associated with each panel device of the plurality of panel devices;

determining scores representing similarities between ones of a first group of census devices determined to have matching panel devices in the panel data and ones of a second group of census devices determined to be unmatched in the panel data; and

assigning each census device of the plurality of census devices a respective demographic class based on the panel demographic information and the scores representing the similarities.

8 . The computing system of claim 7 , wherein determining the scores representing the similarities between the ones of the first group of census devices determined to have the matching panel devices in the panel data and the ones of the second group of census devices determined to be unmatched in the panel data comprises:

implementing a demographic attribution process based on at least one of a recurrent neural network, a Siamese network, or geometric distance calculations to determine the scores representing the similarities.

9 . The computing system of claim 7 , wherein the respective demographic class comprises at least one of age, gender, education, working status, city size, or household size.

10 . The computing system of claim 7 , wherein assigning each census device of the plurality of census devices the respective demographic class comprises applying a machine learning (ML) model trained on the panel demographic information to the census devices.

11 . The computing system of claim 7 , wherein respective numbers of census devices assigned to corresponding ones of the panel devices is in accordance with weights of the corresponding ones of the panel devices.

12 . A computing system comprising:

a processor; and

a non-transitory computer-readable storage medium, having stored thereon program instructions that, upon execution by the processor, cause performance of a set of operations comprising:

accessing census data including census viewing statements associated with media content presented by a plurality of census devices;

accessing panel data including panelist viewing statements associated with media content presented by a plurality of panel devices, the panel data further including panel demographic information associated with each panel device of the plurality of panel devices;

determining scores representing similarities between ones of a first group of census devices determined to have matching panel devices in the panel data and ones of a second group of census devices determined to be unmatched in the panel data; and

assigning each census device of the plurality of census devices a respective demographic class based on the panel demographic information and the scores representing the similarities, wherein respective numbers of census devices assigned to corresponding ones of the panel devices is in accordance with weights of the corresponding ones of the panel devices, wherein the panel data including the weights that are representative of numbers of individuals in a population to be represented by corresponding panelists represented in the panel data, the operations further comprising:

dividing the weights for respective ones of the panelists into respective sets of split weights; and

assigning census viewing statements to at least subsets of the sets of split weights to determine the audience ratings for the population, the assignment of the census viewing statements to the at least subsets of the sets of split weights based on a reach constraint.

13 . The computing system of claim 7 , further comprising:

wherein accessing the panel data comprises:

accessing the panel data from the at least one streaming meter.

14 . The computing system of claim 7 , wherein the digital viewing activity comprises additional digital event information including a platform type and a panelist ID associated with the primary device user for each panel device of the plurality of panel devices.

15 . A non-transitory computer-readable storage medium, having stored thereon program instructions that, upon execution by a processor, cause performance of a set of operations comprising:

accessing census data including census viewing statements associated with media content presented by a plurality of census devices;

accessing panel data including digital viewing activity and panelist viewing statements associated with media content presented by a plurality of panel devices, the panel data further including panel demographic information associated with each panel device of the plurality of panel devices, wherein accessing the panel data comprises:

accessing the panel data from at least one streaming meter configured to measure the digital viewing activity via a broadband connection on each panel device of the plurality of panel devices in a respective broadband capable panel household, wherein the at least one streaming meter is connected to a household WiFi network in the respective broadband capable panel household, wherein each panel device of the plurality of panel devices in the respective broadband capable panel household is registered with a primary device user and mapped to the at least one streaming meter, the digital viewing activity comprising digital event information including the primary device user for each panel device of the plurality of panel devices;

determining scores representing similarities between ones of a first group of census devices determined to have matching panel devices in the panel data and ones of a second group of census devices determined to be unmatched in the panel data; and

assigning each census device of the plurality of census devices a respective demographic class based on the panel demographic information and the scores representing the similarities.

16 . The non-transitory computer-readable storage medium of claim 15 , wherein the operations further comprising:

implementing a demographic attribution process based on at least one of a recurrent neural network, a Siamese network, or geometric distance calculations to determine the scores representing the similarities.

17 . The non-transitory computer-readable storage medium of claim 15 , wherein the respective demographic class comprises at least one of age, gender, education, working status, city size, or household size.

18 . The non-transitory computer-readable storage medium of claim 15 , wherein assigning each census device of the plurality of census devices the respective demographic class comprises applying a machine learning (ML) model trained on the panel demographic information to the census devices.

19 . The non-transitory computer-readable storage medium of claim 15 , wherein respective numbers of census devices assigned to corresponding ones of the panel devices is in accordance with weights of the corresponding ones of the panel devices.

20 . The non-transitory computer-readable storage medium of claim 15 , wherein the digital viewing activity further including information related to a platform associated with the at least one streaming meter and a panelist ID associated with the primary device user for each panel device of the plurality of panel devices.