Telecommunication Network Configuration via Feature-Based Extrapolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for inferring television consumption information across all households, including non-reporting ones, often rely on assumptions that may be incorrect, such as linear scaling or clustering, which lose feature importance and do not accurately represent viewer behavior.
Innovation Solution
A method involving feature-based segmentation and linking, where households are grouped by shared feature values, with non-reporting households linked to reporting households within segments, allowing for accurate extrapolation of television consumption data while retaining variance and importance of features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If linear scaling approach is used to calculate aggregate television consumption information, then the calculation process is simple, but the accuracy of the extrapolation deteriorates because it assumes reporting households represent non-reporting households which may be incorrect
Solution Approach 1:
The patent segments households into distinct groups (reporting households and non-reporting households) rather than treating all households uniformly. This segmentation allows for more accurate extrapolation by recognizing that different household types may have different viewing behaviors, thereby resolving the contradiction between simple calculation and accurate extrapolation.
Solution Approach 2:
The patent changes the extrapolation parameter from a uniform linear scaling factor to segment-specific scaling factors. By applying different scaling approaches to different household segments, the system maintains calculation simplicity while improving extrapolation accuracy for each segment.
2Measurement precision
If clustering approach is used to calculate aggregate television consumption information, then the segmentation of households is improved, but the importance of features is lost making it unclear which features contributed more to clusters
Solution Approach 1:
The patent uses segmentation based on specific household features (such as demographic characteristics, service subscription types, and viewing habits) to create meaningful groups. This approach maintains feature importance by explicitly using features as segmentation criteria rather than creating opaque clusters, thereby resolving the contradiction between segmentation quality and feature information retention.
Solution Approach 2:
The patent changes from cluster-based aggregation to feature-based segmentation with explicit feature weighting. This allows the system to maintain both good segmentation quality and visibility into which features contribute most to the segmentation, eliminating the information loss about feature importance.
3Measurement precision
If feature-based segmentation and linking is used to extrapolate television consumption data, then the accuracy of information extrapolation is improved, but the complexity of the processing system increases
Solution Approach 1:
The patent segments households into reporting and non-reporting groups with distinct processing paths. This segmentation simplifies the overall system complexity by creating manageable sub-systems rather than requiring a single complex extrapolation mechanism for all households.
Solution Approach 2:
The patent uses reporting households as templates or copies to represent non-reporting households within the same segment. By copying data patterns from reporting households to infer non-reporting household behavior, the system achieves accurate extrapolation without requiring complex modeling for each individual household.
Data Source
AI summary
A processing system may obtain a feature set for segmenting households comprising subscribers of a telecommunication network into segments, the households including reporting households for which an information value regarding a feature of interest is available, and non-reporting households for which an information value regarding the feature of interest is not available. The processing system may then assign the households to segments, each segment associated with a set of information values for features of the feature set, and where for each segment, households assigned to the segment have information values that are the same for each of the features of the feature set. The processing system may also link each non-reporting household in a segment to a reporting household in the segment. The processing system may then reconfigure the telecommunication network in accordance with information values for the at least one feature of interest for the plurality of households.