Regional Viewership Prediction via Demand Proxy Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge lies in accurately predicting regional viewership for broadcast media events without direct response variables, which is crucial for resource allocation in industries like food delivery services, as inaccurate predictions can lead to over- or under-allocation of resources.
Innovation Solution
A non-supervised learning approach is implemented, estimating a population cap for each region, deriving a season round function, and using a non-parametric probability framework to generate viewership probabilities based on features and weights of future broadcast events, leveraging national viewership or demand data to learn relationships between historical events.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional supervised learning is used for viewership prediction, then the model can learn from direct response variables, but regional viewership data is unavailable making this approach inapplicable
Solution Approach 1:
The patent uses demand data for goods and services as an intermediary variable to proxy for viewership information. Instead of directly measuring viewership (which is unavailable), the system measures demand for complementary goods/services during broadcast events and uses this as a surrogate signal to train the prediction model.
Solution Approach 2:
The patent replaces the conventional supervised learning mechanism (which requires direct response variables) with an unsupervised learning approach that infers relationships from auxiliary data. The mechanical system of direct measurement is substituted with an inferential system using demand data as a proxy.
2Adaptability or versatility
If predictions are made without direct regional viewership data, then the model must use alternative data sources, but this increases model complexity and difficulty of learning relationships
Solution Approach 1:
The patent segments the prediction problem into distinct components: (1) identifying broadcast events, (2) measuring demand for goods/services during these events, (3) extracting features from demand data, and (4) training the prediction model. This segmentation makes the complex task of predicting without direct data more manageable.
Solution Approach 2:
The patent changes the parameters used for training by substituting viewership counts with demand metrics for goods and services. By transforming the response variable from direct viewership measurement to indirect demand measurement, the model adapts to work with available data while maintaining predictive capability.
3Reliability
If inaccurate predictions are made, then resource allocation becomes unreliable, but improving accuracy requires better data that is not available
Solution Approach 1:
The patent implements a feedback mechanism where demand data from actual broadcast events is continuously collected and used to refine the prediction model. The system learns from historical demand patterns during televised events and improves its predictions over time, creating a closed-loop system that enhances reliability.
Solution Approach 2:
The system uses its own collected demand data to improve its predictions without requiring external viewership data. The model serves itself by learning from the patterns it observes in demand data, making the system self-improving and reducing dependence on unavailable direct measurements.
Data Source
AI summary
Techniques for regional viewership predictions of broadcast events such as live broadcast professional sporting events. The techniques can make the predictions without a direct response variable such as regional viewership data for training a prediction model. Instead, in one technique, demand information for a good or service is used. From the demand information, a derivative demand for the good or service relative to a normal demand is determined. A regression framework is used to learn relationships between the derivative demand for the good or service and features of past live broadcast sporting events. This results in a matrix of feature weights. A non-parametric mixture framework is then used to find a set of feature weights that can be applied to features of future broadcast events to generate regional viewership predictions for the events.