Clustering television programs based on viewing behavior
View Patent ↗Example program clustering methods disclosed herein include accessing person-level program viewing data representing lengths of time respective people in an audience have tuned to respective television programs to be clustered. Disclosed example methods also include determining adjusted person-level program viewing data for respective ones of the people having tuned to respective ones of the television programs. For example, first person-level program viewing data for a first person having tuned to a first program is adjusted based on a ratio characterizing a relationship between a first program rating associated with the first person having tuned to the first program and a first network rating associated with the first person having tuned to a first network associated with the first program. Disclosed example methods further include clustering the television programs into clusters based on distances between pairs of the television programs, the distances based on the adjusted person-level program viewing data.
1. An apparatus comprising:
memory including computer readable instructions;
a processor to execute the computer readable instructions;
a program viewing adjuster implemented by the processor to:
access respective person-level program viewing data that is to represent lengths of time respective people in an audience have tuned to respective ones of a plurality of television programs to be clustered; and
determine adjusted person-level program viewing data for respective ones of the people having tuned to respective ones of the television programs, first person-level program viewing data for a first one of the people having tuned to a first one of the programs to be adjusted by the program viewing adjuster based on a first ratio that is to characterize a relationship between (i) a first program rating associated with the first one of the people having tuned to the first one of the programs and (ii) a first network rating associated with the first one of the people having tuned to a first network associated with the first one of the programs, wherein to determine the first ratio, the program viewing adjuster is to:
determine a numerator value based on a difference between the first program rating and the first network rating;
determine a denominator value based on the first network rating and a total duration of the first one of the programs; and
divide the numerator value by the denominator value to determine the first ratio; and
a program clusterer implemented by the processor to cluster the ones of the plurality of television programs into clusters based on distances between pairs of the television programs, the distances based on the adjusted person-level program viewing data.
2. The apparatus of claim 1 , wherein the first program rating corresponds to a fraction of time, relative to a total duration of the first one of the programs, during which the first one of the people tuned to the first one of the programs, and the first network rating corresponds to a fraction of time, relative to a monitoring interval, during which the first one of the people tuned to the first network.
3. The apparatus of claim 1 , wherein the program viewing adjuster is to adjust the first person-level program viewing data for the first one of the people having tuned to the first one of the programs by:
comparing the first ratio to a threshold;
setting the first person-level program viewing data equal to zero when the ratio does not satisfy the threshold; and
leaving the first person-level program viewing data unchanged when the ratio satisfies the threshold.
4. The apparatus of claim 3 , wherein the threshold is satisfied when the ratio exceeds the threshold, and the threshold is not satisfied when the ratio does not exceed the threshold.
5. An apparatus comprising:
memory including computer readable instructions;
a processor to execute the computer readable instructions;
a program viewing adjuster implemented by the processor to:
access respective person-level program viewing data that is to represent lengths of time respective people in an audience have tuned to respective ones of a plurality of television programs to be clustered; and
determine adjusted person-level program viewing data for respective ones of the people having tuned to respective ones of the television programs, first person-level program viewing data for a first one of the people having tuned to a first one of the programs to be adjusted by the program viewing adjuster based on a first ratio that is to characterize a relationship between (i) a first program rating associated with the first one of the people having tuned to the first one of the programs and (ii) a first network rating associated with the first one of the people having tuned to a first network associated with the first one of the programs;
a program clusterer implemented by the processor to cluster the ones of the plurality of television programs into clusters based on distances between pairs of the television programs, the distances to be based on the adjusted person-level program viewing data; and
a cluster number selector implemented by the processor to select a final number of clusters into which the plurality of television programs is to be clustered, the cluster number selector to select the final number of clusters based on average silhouette width values determined for clustering the plurality of television programs into different possible numbers of clusters.
6. The apparatus of claim 5 , wherein the cluster number selector is to determine an average silhouette width value for clustering the plurality of television programs into a first possible number of clusters by:
computing silhouette width values for respective ones of the plurality of television programs when the television programs are clustered into the first possible number of clusters, a silhouette width value for a first one of the television programs based on (1) an average distance of the first one of the television programs to other television programs in a same first cluster as the first one of the television programs, and (2) an average distance of the first one of the television programs to a neighboring cluster of the first cluster; and
averaging the silhouette width values for the respective ones of the plurality of television programs when the television programs are clustered into the first possible number of clusters to determine the average silhouette width value for clustering the plurality of television programs into the first possible number of clusters.
7. A non-transitory computer readable storage medium comprising computer readable instructions which, when executed, cause a processor to at least:
access respective person-level program viewing data that is to represent lengths of time respective people in an audience have tuned to respective ones of a plurality of television programs to be clustered;
determine adjusted person-level program viewing data for respective ones of the people having tuned to respective ones of the television programs, first person-level program viewing data for a first one of the people having tuned to a first one of the programs to be adjusted based on a first ratio that is to characterize a relationship between (i) a first program rating associated with the first one of the people having tuned to the first one of the programs and (ii) a first network rating associated with the first one of the people having tuned to a first network associated with the first one of the programs; and
cluster the ones of the plurality of television programs into clusters based on distances between pairs of the television programs, the distances being based on the adjusted person-level program viewing data, wherein to determine the first ratio, the instructions, when executed, cause the processor to:
determine a numerator value based on a difference between the first program rating and the first network rating;
determine a denominator value based on the first network rating and a total duration of the first one of the programs; and
divide the numerator value by the denominator value to determine the first ratio.
8. The storage medium of claim 7 , wherein the first program rating corresponds to a fraction of time, relative to a total duration of the first one of the programs, during which the first one of the people tuned to the first one of the programs, and the first network rating corresponds to a fraction of time, relative to a monitoring interval, during which the first one of the people tuned to the first network.
9. The storage medium of claim 7 , wherein the instructions, when executed, cause the processor to adjust the first person-level program viewing data for the first one of the people having tuned to the first one of the programs by:
comparing the first ratio to a threshold;
setting the first person-level program viewing data equal to zero when the ratio does not satisfy the threshold; and
leaving the first person-level program viewing data unchanged when the ratio satisfies the threshold.
10. A non-transitory computer readable storage medium comprising computer readable instructions which, when executed, cause a processor to at least:
access respective person-level program viewing data that is to represent lengths of time respective people in an audience have tuned to respective ones of a plurality of television programs to be clustered;
determine adjusted person-level program viewing data for respective ones of the people having tuned to respective ones of the television programs, first person-level program viewing data for a first one of the people having tuned to a first one of the programs to be adjusted based on a first ratio that is to characterize a relationship between (i) a first program rating associated with the first one of the people having tuned to the first one of the programs and (ii) a first network rating associated with the first one of the people having tuned to a first network associated with the first one of the programs;
cluster the ones of the plurality of television programs into clusters based on distances between pairs of the television programs, the distances being based on the adjusted person-level program viewing data; and
select a final number of clusters into which the plurality of television programs is to be clustered, the final number of clusters to be selected based on average silhouette width values determined for clustering the plurality of television programs into different possible numbers of clusters.
11. The storage medium of claim 10 , wherein the instructions, when executed, cause the processor to determine an average silhouette width value for clustering the plurality of television programs into a first possible number of clusters by:
computing silhouette width values for respective ones of the plurality of television programs when the television programs are clustered into the first possible number of clusters, a silhouette width value for a first one of the television programs based on (1) an average distance of the first one of the television programs to other television programs in a same first cluster as the first one of the television programs, and (2) an average distance of the first one of the television programs to a neighboring cluster of the first cluster; and
averaging the silhouette width values for the respective ones of the plurality of television programs when the television programs are clustered into the first possible number of clusters to determine the average silhouette width value for clustering the plurality of television programs into the first possible number of clusters.
12. A program clustering method comprising:
accessing, by executing an instruction with a processor, respective person-level program viewing data representing lengths of time respective people in an audience have tuned to respective ones of a plurality of television programs to be clustered;
determining, by executing an instruction with the processor, adjusted person-level program viewing data for respective ones of the people having tuned to respective ones of the television programs, first person-level program viewing data for a first one of the people having tuned to a first one of the programs being adjusted based on a first ratio characterizing a relationship between (i) a first program rating associated with the first one of the people having tuned to the first one of the programs and (ii) a first network rating associated with the first one of the people having tuned to a first network associated with the first one of the programs;
determining, by executing an instruction with the processor, the first ratio by:
determining a numerator value based on a difference between the first program rating and the first network rating;
determining a denominator value based on the first network rating and a total duration of the first one of the programs; and
dividing the numerator value by the denominator value to determine the first ratio; and
clustering, by executing an instruction with the processor, the ones of the plurality of television programs into clusters based on distances between pairs of the television programs, the distances being based on the adjusted person-level program viewing data.
13. The method of claim 12 , wherein the first program rating corresponds to a fraction of time, relative to a total duration of the first one of the programs, during which the first one of the people tuned to the first one of the programs, and the first network rating corresponds to a fraction of time, relative to a monitoring interval, during which the first one of the people tuned to the first network.
14. The method of claim 12 , wherein the determining of the adjusted person-level program viewing data includes adjusting the first person-level program viewing data for the first one of the people having tuned to the first one of the programs by:
comparing the first ratio to a threshold;
setting the first person-level program viewing data equal to zero when the ratio does not satisfy the threshold; and
leaving the first person-level program viewing data unchanged when the ratio satisfies the threshold.
15. The method of claim 14 , wherein the threshold is satisfied when the ratio exceeds the threshold, and the threshold is not satisfied when the ratio does not exceed the threshold.
16. A program clustering method comprising:
accessing, by executing an instruction with a processor, respective person-level program viewing data representing lengths of time respective people in an audience have tuned to respective ones of a plurality of television programs to be clustered;
determining, by executing an instruction with the processor, adjusted person-level program viewing data for respective ones of the people having tuned to respective ones of the television programs, first person-level program viewing data for a first one of the people having tuned to a first one of the programs being adjusted based on a first ratio characterizing a relationship between (i) a first program rating associated with the first one of the people having tuned to the first one of the programs and (ii) a first network rating associated with the first one of the people having tuned to a first network associated with the first one of the programs;
clustering, by executing an instruction with the processor, the ones of the plurality of television programs into clusters based on distances between pairs of the television programs, the distances being based on the adjusted person-level program viewing data; and
selecting, by executing an instruction with the processor, a final number of clusters into which the plurality of television programs is to be clustered, the selecting based on determining average silhouette width values for clustering the plurality of television programs into different possible numbers of clusters.
17. The method of claim 16 , wherein an average silhouette width value is determined for clustering the plurality of television programs into a first possible number of clusters by:
computing silhouette width values for respective ones of the plurality of television programs when the television programs are clustered into the first possible number of clusters, a silhouette width value for a first one of the television programs being based on (1) an average distance of the first one of the television programs to other television programs in a same first cluster as the first one of the television programs, and (2) an average distance of the first one of the television programs to a neighboring cluster of the first cluster; and
averaging the silhouette width values for the respective ones of the plurality of television programs when the television programs are clustered into the first possible number of clusters to determine the average silhouette width value for clustering the plurality of television programs into the first possible number of clusters.