Methods and apparatus to categorize media impressions by age
An example apparatus includes a memory management unit (MMU) to store audience member records in a first block of memory, the audience member records including attribute-value pairs representative of database subscriber activity data of corresponding audience members subscribed to a database proprietor. The MMU to split the audience member records from an initial node into child nodes based on comparisons of first ones of the attribute-value pairs of the audience member records to a first value threshold. When an arithmetic logic unit (ALU) determines that a quantity of ones of the audience member records does not satisfy a minimum leaf size, store a terminal node value in the second block of the memory to indicate the first child node is associated with one age category. When the ALU determines that the quantity of the ones of the audience member records satisfies the minimum leaf size, split the first child node into second child nodes, the audience member records in the first child node to be: (1) assigned to corresponding ones of the second child nodes based on comparisons of second ones of the attribute-value pairs of the audience member records to a second value threshold, and (2) processed further to create terminal nodes. Generate an age-correction model based on the terminal nodes to facilitate correcting a database subscriber age characteristic associated with a media impression that is logged by a server of the database proprietor based on a network communication transmitted by a client device to report to the server of the database proprietor that media was accessed via the client device.
1. A processor system, comprising:
memory in circuit with a processor;
a memory management unit (MMU) to:
store audience member records in a first block of the memory, the audience member records including attribute-value pairs representative of database subscriber activity data of corresponding audience members subscribed to a database proprietor;
split the audience member records from an initial node into child nodes based on comparisons of first ones of the attribute-value pairs of the audience member records to a first value threshold;
load, into a cache memory of the processor, a portion of the audience member records corresponding to a first one of the child nodes stored in a second block of the memory; and
an arithmetic logic unit (ALU) to:
compare first and second registers of the processor, the first register including a quantity of ones of the audience member records that are in the first child node, and the second register including a minimum leaf size;
the MMU to:
when the ALU determines that the quantity of the ones of the audience member records in the first child node does not satisfy the minimum leaf size, store a terminal node value in the second block of the memory to indicate the first child node as a terminal node associated with one age category;
when the ALU determines that the quantity of the ones of the audience member records in the first child node satisfies the minimum leaf size, store an intermediate node value in the second block of the memory to indicate the first child node as an intermediate node, and split the first child node into second child nodes to be stored at corresponding third blocks of the memory, the ones of the audience member records in the first child node to be: (1) assigned to corresponding ones of the second child nodes based on comparisons of second ones of the attribute-value pairs of the audience member records to a second value threshold, and (2) processed further to create terminal nodes; and
generate an age-correction model based on the terminal nodes to facilitate correcting a database subscriber age characteristic associated with a media impression that is logged by a server of the database proprietor based on a network communication transmitted by a client device to report to the server of the database proprietor that media was accessed via the client device.
2. The processor system as defined in claim 1 , wherein the MMU is to split the audience member records into the child nodes based on comparisons performed by the ALU between the first threshold value in a third register and values of the first attribute-value pairs in fourth registers.
3. The processor system as defined in claim 1 , wherein the MMU is further to assign a weight to each audience member record, the weight being based on a quantity of audience members in a same age group as the audience member record.
4. The processor system as defined in claim 3 , wherein the MMU is to store an array of age categories in a fourth block of the memory and to load a fifth register with a value of a position in the array of age categories, when the value in the fifth register equals the last position in the array of age categories, the MMU is to determine that all the weights have been assigned.
5. The processor system as defined in claim 1 , wherein the MMU is further to organize the audience member records into a training set and a validation set, the training set to be stored in the first block of the memory and the validation set to be stored in a fourth block of the memory, the ALU to apply the audience member records in the validation set to the age correction model to validate the age correction model.
6. The processor system as defined in claim 5 , wherein the ALU is to apply the audience member records in the validation set to the age correction model by calculating a difference between registers containing values of an expected output of the age correction model and registers containing values of an actual output of the age correction model.
7. A non-transitory computer readable storage medium comprising machine readable instructions that, when executed, cause a processor system to at least:
store audience member records in a first block of memory, the audience member records including attribute-value pairs representative of database subscriber activity data of corresponding audience members subscribed to a database proprietor;
split the audience member records from an initial node into child nodes based on comparisons of first ones of the attribute-value pairs of the audience member records to a first value threshold;
load, into a cache memory of a processor of the processor system, a portion of the audience member records corresponding to a first one of the child nodes stored in a second block of the memory; and
compare, with an arithmetic logic unit (ALU) of the processor system, first and second registers of the processor, the first register including a quantity of ones of the audience member records that are in the first child node, and the second register including a minimum leaf size;
when the quantity of the ones of the audience member records in the first child node does not satisfy the minimum leaf size, store a terminal node value in the second block of the memory to indicate the first child node as a terminal node associated with one age category;
when the quantity of the ones of the audience member records in the first child node satisfies the minimum leaf size, store an intermediate node value in the second block of the memory to indicate the first child node as an intermediate node, and split the first child node into second child nodes to be stored at corresponding third blocks of the memory, the ones of the audience member records in the first child node to be: (1) assigned to corresponding ones of the second child nodes based on comparisons of second ones of the attribute-value pairs of the audience member records to a second value threshold, and (2) processed further to create terminal nodes; and
generate an age-correction model based on the terminal nodes to facilitate correcting a database subscriber age characteristic associated with a media impression that is logged by a server of the database proprietor based on a network communication transmitted by a client device to report to the server of the database proprietor that media was accessed via the client device.
8. The non-transitory computer readable storage medium as defined in claim 7 , further including instructions that, when execute cause the processor system to split the audience member records into the child nodes based on comparisons performed by the ALU between the first threshold value in a third register and values of the first attribute-value pairs in fourth registers.
9. The non-transitory computer readable storage medium as defined in claim 7 , further including instructions that, when execute cause the processor system to assign a weight to each audience member record, the weight being based on a quantity of audience members in a same age group as the audience member record.
10. The non-transitory computer readable storage medium as defined in claim 9 , further including instructions that, when executed cause the processor system to store an array of age categories in a third block of the memory and to load a fifth register with a value of a position in the array of age categories, when the value in the fifth register equals the last position in the array of age categories, determine that all the weights have been assigned.
11. The non-transitory computer readable storage medium as defined in claim 7 , further including instructions that, when executed cause the processor system to organize the audience member records into a training set and a validation set, the training set to be stored in the first block of the memory and the validation set to be stored in a fourth block of the memory, apply, with the ALU, the audience member records in the validation set to the age correction model to validate the age correction model.
12. The non-transitory computer readable storage medium as defined in claim 11 , wherein the audience member records in the validation set are applied to the age correction model by calculating, using the ALU, a difference between registers containing values of an expected output of the age correction model and registers containing values of an actual output of the age correction model.