Household Classification Using Similarity-Based Viewing Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Household addressable media networks face challenges in determining which segment(s) certain households belong to due to a lack of available data, making it difficult to target specific households with addressable supplemental content.
Innovation Solution
A system and method to determine household segments by comparing known households with available data to unknown households, using features and behaviors to identify similarities, and adjusting data to accurately predict content viewing behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If household addressable media networks attempt to target specific household segments, then content delivery precision is improved, but data availability deteriorates because the network does not have access to data associated with a large number of households
Solution Approach 1:
The patent introduces third-party data sources as intermediaries that bridge the gap between the media network and household data. These external sources provide demographic, geographic, and behavioral data that the media network cannot obtain directly, enabling accurate household classification without requiring direct access to household information
Solution Approach 2:
The patent replaces direct data collection mechanisms with statistical inference and machine learning models. Instead of mechanically collecting data from each household, the system uses algorithms to predict household characteristics and segment memberships based on available proxy data and patterns from similar households
2Measurement precision
If the network collects more household data to improve segmentation, then classification accuracy is improved, but system complexity increases
Solution Approach 1:
The patent segments the household classification problem into multiple independent stages: data collection from various sources, feature extraction, similarity calculation, and segment assignment. This modular approach reduces overall system complexity by breaking down the complex task into manageable, independent components that can be processed separately
Solution Approach 2:
The patent creates a universal classification framework that handles multiple data types (demographic, geographic, behavioral) and multiple segment categories through a single integrated system. This multi-functional approach avoids the need for separate specialized systems for each data type or segment, reducing overall complexity
Data Source
AI summary
A group of households associated with known content viewing behavior may be determined. The group of households may be compared to at least one household with unknown content viewing behavior, based on a plurality of characteristics. Based on the comparison, a subset of the group of households that is associated with characteristics most similar to characteristics associated with the at least one household with unknown viewing behavior may be determined. A predicted content viewing behavior associated with the at least one household with unknown content viewing behavior may be determined based on the known content viewing behavior of the subset of the group of households.