Methods and apparatus to project ratings for future broadcasts of media
View Patent ↗Methods, apparatus, systems and articles of manufacture are disclosed to project ratings for future broadcasts of media. Disclosed example methods include normalizing, with a processor, audience measurement data corresponding to media exposure data, social media exposure data and programming information associated with a future quarter to determine normalized audience measurement data. Disclosed example methods also include classifying a media asset based on the programming information to determine a media asset classification. Disclosed example methods also include building, with the processor, a projection model based on a first subset of the normalized audience measurement data, the first subset of the normalized audience measurement data associated with a first time frame relative to the future quarter, the first subset of the normalized audience measurement data based on the media asset classification, and applying, with the processor, the programming information to the projection model to project ratings for the media asset.
1. An upfront ratings projector apparatus comprising:
a data transformer to transform audience measurement data to determine normalized training data having a common scale, the audience measurement data including media exposure data, social media exposure data and programming information associated with a media asset for which ratings are to be predicted for a future quarter of programming;
a model builder to:
select one of a first machine learning model or a second machine learning model to build based on the future quarter, the model builder to exclude historical data based on a gap between a current quarter and the future quarter when building the second machine learning model but not when building the first machine learning model,
select a first subset of predictive features from the normalized training data according to a predictive feature schema, the model builder to select the predictive feature schema from a plurality of predictive feature schemas based on a classification of the media asset, the first subset of predictive features including historical ratings data for the media asset,
select a second subset of predictive features from the normalized training data to train the other of the selected one of the first machine learning model or the second machine learning model, the second subset of predictive features selected from a predictive feature schema different than the predictive feature schema used to select the first subset of predictive features,
train the selected one of the first or second machine learning model based on at least a portion of the first subset of predictive features to reduce error between the historical ratings data and predicted ratings data output by the selected one of the first or second machine learning model based on the first subset of predictive features, and
a ratings projector to apply the selected one of the first or second machine learning model to predict ratings for the media asset for the future quarter, at least one of the data transformer, the model builder or the ratings projector implemented by a logic circuit.
2. The apparatus as defined in claim 1 , wherein the model builder is to classify the media asset as a television series when a characteristic of the media asset is indicative of at least one of a premier episode, a repeat episode or a new episode.
3. The apparatus as defined in claim 2 , wherein the model builder is to retrieve series historical performance information when the media asset is classified as a television series.
4. The apparatus as defined in claim 1 , wherein the model builder is to classify the media asset as special programming when a characteristic of the media asset is indicative of at least one of a movie or a sporting event.
5. The apparatus as defined in claim 1 , wherein the first subset of predictive features includes historical ratings data for (i) the media asset and (ii) a subset of media assets that are (1) included in the normalized training data and (2) related to the media asset.
6. A tangible computer-readable storage medium comprising instructions that, when executed, cause a processor to at least:
transform audience measurement data to determine normalized training data having a common scale, the audience measurement data including media exposure data, social media exposure data and programming information associated with a media asset for which ratings are to be predicted for a future quarter of programming;
select one of a first machine learning model or a second machine learning model to build based on the future quarter, historical data to be excluded based on a gap between a current quarter and the future quarter when building the second machine learning model but not when building the first machine learning model;
select a first subset of predictive features from the normalized training data according to a predictive feature schema, the predictive feature schema selected from a plurality of predictive feature schemas based on a classification of the media asset, the first subset of predictive features including historical ratings data for the media asset;
select a second subset of predictive features from the normalized training data to train the other of the selected one of the first machine learning model or the second machine learning model, the second subset of predictive features selected from a predictive feature schema different than the predictive feature schema used to select the first subset of predictive features;
train the selected one of the first or second machine learning model based on at least a portion of the first subset of predictive features to reduce error between the historical ratings data and predicted ratings data output by the selected one of the first or second machine learning model based on the first subset of predictive features; and
apply the selected one of the first or second machine learning model to predict ratings for the media asset for the future quarter.
7. The tangible machine-readable storage medium as defined in claim 6 , wherein the instructions further cause the processor to:
classify the media asset as a television series when a characteristic of the media asset is indicative of at least one of a premier episode, a repeat episode or a new episode; and
retrieve series historical performance information related to the media asset when the media asset is classified as a television series.
8. The tangible machine-readable storage medium as defined in claim 7 , wherein the instructions further cause the processor to classify the media asset as special programming when a characteristic of the media asset is indicative of a movie or a sporting event.
9. The tangible machine-readable storage medium as defined in claim 6 , wherein the first subset of predictive features includes historical ratings data for (i) the media asset and (ii) a subset of media assets that are (1) included in the normalized training data and (2) related to the media asset.
10. An upfront ratings projection method comprising:
transforming, by executing an instruction with a processor, audience measurement data to determine normalized training data having a common scale, the audience measurement data including media exposure data, social media exposure data and programming information associated with a media asset for which ratings are to be predicted for a future quarter of programming;
selecting, by executing an instruction with the processor, one of a first machine learning model or a second machine learning model to build based on the future quarter, historical data to be excluded based on a gap between a current quarter and the future quarter when building the second machine learning model but not when building the first machine learning model,
selecting, by executing an instruction with the processor, a first subset of predictive features from the normalized training data according to a predictive feature schema, the predictive feature schema to be selected from a plurality of predictive feature schemas based on a classification of the media asset, the first subset of predictive features including historical ratings data for the media asset,
selecting, by executing an instruction with the processor, a second subset of predictive features from the normalized training data to train the other of the selected one of the first machine learning model or the second machine learning model, the second subset of predictive features selected from a predictive feature schema different than the predictive feature schema used to select the first subset of predictive features;
training, by executing an instruction with the processor, the selected one of the first or second machine learning model based on at least a portion of the first subset of predictive features to reduce error between the historical ratings data and predicted ratings data output by the selected one of the first or second machine learning model based on the first subset of predictive features; and
applying, by executing an instruction with the processor, the selected one of the first or second machine learning model to predict ratings for the media asset for the future quarter.
11. The method as defined in claim 10 , wherein the classifying further includes classifying the media asset as a television series when a characteristic of the media asset is indicative of at least one of a premier episode, a repeat episode or a new episode.
12. The method as defined in claim 11 , further including, in response to the classifying of the media asset as a television series, retrieving series historical performance information related to the media asset.
13. The method as defined in claim 10 , wherein the classifying further includes classifying the media asset as special programming when a characteristic of the media asset is indicative of at least one of a movie or a sporting event.
14. The method as defined in claim 10 , wherein the first subset of predictive features includes historical ratings data for (i) the media asset and (ii) a subset of media assets that are (1) included in the normalized training data and (2) related to the media asset.