Methods and apparatus to perform identity matching across audience measurement systems
View Patent ↗Methods, apparatus, systems and articles of manufacture to perform identity matching across audience measurement systems are disclosed. An example apparatus includes a data normalizer to normalize audience measurement events corresponding to media exposure data obtained from a first audience measurement system and a second audience measurement system, a tree builder to build a k-dimensional tree based on normalized audience measurement events, a candidate identifier to calculate a distance between a query event and a first audience measurement event included in a search space in the k-dimensional tree, identify the query event and the first audience measurement event as a candidate match when the calculated distance satisfies a distance threshold, an array builder to generate metrics at an identifier-level based on a plurality of candidate matches, and a thresholder to identify an identifier mapping linking a first user identifier to a second user identifier when the metrics satisfy respective metric thresholds.
1. An apparatus to perform identity matching, the apparatus comprising:
a data normalizer to normalize audience measurement events corresponding to media exposure data obtained from a first audience measurement system and a second audience measurement system;
a tree builder to build a k-dimensional tree based on normalized audience measurement events corresponding to the first audience measurement system, the k-dimensional tree including a plurality of search spaces;
a candidate identifier to:
identify a search space in the k-dimensional tree based on a query event corresponding to the second audience measurement system;
calculate a distance between the query event and a first audience measurement event included in the search space, the first audience measurement event corresponding to the first audience measurement system; and
identify the query event and the first audience measurement event as a candidate match when the calculated distance satisfies a distance threshold;
an array builder to generate metrics at an identifier-level based on a plurality of candidate matches; and
a thresholder to identify an identifier mapping linking a first user identifier associated with the first audience measurement system to a second user identifier associated with the second audience measurement system when the metrics satisfy respective metric thresholds, wherein at least one of the data normalizer, the tree builder, the candidate identifier, the array builder or the thresholder is implemented by at least one of hardware or at least one processor.
2. The apparatus as defined in claim 1 , further including a matrix generator to:
generate a first matrix based on normalized audience measurement events corresponding to the first audience measurement system; and
generate a second matrix based on normalized audience measurement events corresponding to the second audience measurement system.
3. The apparatus as defined in claim 2 , wherein a size associated with the first matrix corresponds to (1) a number of normalized audience measurement events corresponding to the first audience measurement system and (2) a number of variables in common between the normalized audience measurement events corresponding to the first audience measurement system and the normalized audience measurement events corresponding to the second audience measurement system.
4. The apparatus as defined in claim 3 , wherein the tree builder is to:
compare the size associated with the first matrix to a size associated with the second matrix; and
determine the size associated with the first matrix is greater than the size associated with the second matrix when building the k-dimensional tree based on the normalized measurement events corresponding to the first audience measurement system.
5. The apparatus as defined in claim 1 , wherein the array builder is to:
build a first array based on a number of matched events between a first set of user identifiers associated with the first audience measurement system and a second set of user identifiers associated with the second audience measurement system, the first set of user identifiers and the second set of user identifiers included in the plurality of candidate matches;
build a second array based on a first percentage of matched events associated with the first set of user identifiers;
build a third array based on a second percentage of matched events associated with the second set of user identifiers; and
build a fourth array based on clock offsets associated with the plurality of candidate matches.
6. The apparatus as defined in claim 5 , wherein the thresholder is to:
compare a first value associated with an identifier combination from the first array to a first threshold, the identifier combination (1) selected from the first set of user identifiers and the second set of user identifiers and (2) associated with a non-zero cell in the first array;
compare a first value associated with the identifier combination from the first array to a first threshold;
compare a second value associated with the identifier combination from the second array to a second threshold when the first value satisfies the first threshold;
compare a third value associated with the identifier combination from the third array to a third threshold when the second value satisfies the second threshold; and
compare a fourth value associated with the identifier combination from the fourth array to a fourth threshold when the third value satisfies the third threshold.
7. The apparatus as defined in claim 6 , wherein the thresholder is to record the identifier combination as the identifier mapping when the fourth value satisfies the fourth threshold.
8. A method to perform identity matching, the method comprising:
normalizing, by executing an instruction with a processor, audience measurement events corresponding to media exposure data obtained from a first audience measurement system and a second audience measurement system;
building, by executing an instruction with the processor, a k-dimensional tree based on normalized audience measurement events corresponding to the first audience measurement system, the k-dimensional tree including a plurality of search spaces;
identifying, by executing an instruction with the processor, a search space in the k-dimensional tree based on a query event corresponding to the second audience measurement system;
calculating, by executing an instruction with the processor, a distance between the query event and a first audience measurement event included in the search space, the first audience measurement event corresponding to the first audience measurement system;
identifying, by executing an instruction with the processor, the query event and the first audience measurement event as a candidate match when the calculated distance satisfies a distance threshold;
generating, by executing an instruction with the processor, metrics at an identifier-level based on a plurality of candidate matches; and
identifying, by executing an instruction with the processor, an identifier mapping linking a first user identifier associated with the first audience measurement system to a second user identifier associated with the second audience measurement system when the metrics satisfy respective metric thresholds.
9. The method as defined in claim 8 , further including:
generating a first matrix based on normalized audience measurement events corresponding to the first audience measurement system; and
generating a second matrix based on normalized audience measurement events corresponding to the second audience measurement system.
10. The method as defined in claim 9 , wherein a size associated with the first matrix corresponds to (1) a number of normalized audience measurement events corresponding to the first audience measurement system and (2) a number of variables in common between the normalized audience measurement events corresponding to the first audience measurement system and the normalized audience measurement events corresponding to the second audience measurement system.
11. The method as defined in claim 10 , wherein the building of the k-dimensional tree based on the normalized measurement events corresponding to the first audience measurement system includes:
comparing the size associated with the first matrix to a size associated with the second matrix; and
determining the size associated with the first matrix is greater than the size associated with the second matrix.
12. The method as defined in claim 8 , further including:
building a first array based on a number of matched events between a first set of user identifiers associated with the first audience measurement system and a second set of user identifiers associated with the second audience measurement system, the first set of user identifiers and the second set of user identifiers included in the plurality of candidate matches;
building a second array based on a first percentage of matched events associated with the first set of user identifiers;
building a third array based on a second percentage of matched events associated with the second set of user identifiers; and
building a fourth array based on clock offsets associated with the plurality of candidate matches.
13. The method as defined in claim 12 , further including:
comparing a first value associated with an identifier combination from the first array to a first threshold, the identifier combination (1) selected from the first set of user identifiers and the second set of user identifiers and (2) associated with a non-zero cell in the first array;
in response to determining that the first value satisfies the first threshold, comparing a second value associated with the identifier combination from the second array to a second threshold;
in response to determining that the second value satisfies the second threshold, comparing a third value associated with the identifier combination from the third array to a third threshold; and
in response to determining that that the third value satisfies the third threshold, comparing a fourth value associated with the identifier combination from the fourth array to a fourth threshold.
14. The method as defined in claim 13 , further including recording the identifier combination as the identifier mapping in response to determining that the fourth value satisfies the fourth threshold.
15. A tangible computer readable storage medium comprising instructions that, when executed, cause a machine to at least:
normalize audience measurement events corresponding to media exposure data obtained from a first audience measurement system and a second audience measurement system;
build k-dimensional tree based on normalized audience measurement events corresponding to the first audience measurement system, the k-dimensional tree including a plurality of search spaces;
identify a search space in the k-dimensional tree based on a query event corresponding to the second audience measurement system;
calculate a distance between the query event and a first audience measurement event included in the search space, the first audience measurement event corresponding to the first audience measurement system;
identify the query event and the first audience measurement event as a candidate match when the calculated distance satisfies a distance threshold;
generate metrics at an identifier-level based on a plurality of candidate matches; and
identify an identifier mapping linking a first user identifier associated with the first audience measurement system to a second user identifier associated with the second audience measurement system when the metrics satisfy respective metric thresholds.
16. The tangible computer readable storage medium as defined in claim 15 , further including instructions that, when executed, cause the machine to:
generate a first matrix based on normalized audience measurement events corresponding to the first audience measurement system; and
generate a second matrix based on normalized audience measurement events corresponding to the second audience measurement system.
17. The tangible computer readable storage medium as defined in claim 16 , wherein a size associated with the first matrix corresponds to (1) a number of normalized audience measurement events corresponding to the first audience measurement system and (2) a number of variables in common between the normalized audience measurement events corresponding to the first audience measurement system and the normalized audience measurement events corresponding to the second audience measurement system.
18. The tangible computer readable storage medium as defined in claim 17 , further including instructions that, when executed, cause the machine to:
compare the size associated with the first matrix to a size associated with the second matrix; and
determine the size associated with the first matrix is greater than the size associated with the second matrix when building the k-dimensional tree based on the normalized measurement events corresponding to the first audience measurement system.
19. The tangible computer readable storage medium as defined in claim 15 , further including instructions that, when executed, cause the machine to:
build a first array based on a number of matched events between a first set of user identifiers associated with the first audience measurement system and a second set of user identifiers associated with the second audience measurement system, the first set of user identifiers and the second set of user identifiers included in the plurality of candidate matches;
build a second array based on a first percentage of matched events associated with the first set of user identifiers;
build a third array based on a second percentage of matched events associated with the second set of user identifiers; and
build a fourth array based on clock offsets associated with the plurality of candidate matches.
20. The tangible computer readable storage medium as defined in claim 19 , further including instructions that, when executed, cause the machine to:
compare a first value associated with an identifier combination from the first array to a first threshold, the identifier combination (1) selected from the first set of user identifiers and the second set of user identifiers and (2) associated with a non-zero cell in the first array;
compare a first value associated with the identifier combination from the first array to a first threshold;
compare a second value associated with the identifier combination from the second array to a second threshold when the first value satisfies the first threshold;
compare a third value associated with the identifier combination from the third array to a third threshold when the second value satisfies the second threshold;
compare a fourth value associated with the identifier combination from the fourth array to a fourth threshold when the third value satisfies the third threshold; and
record the identifier combination as the identifier mapping when the fourth value satisfies the fourth threshold.