Audience Identity Matching via K-Dimensional Tree Event Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audience measurement systems face challenges in accurately matching identity across disparate data sets from different measurement systems, such as television and subscription provider data, due to asynchronous clocks, differences in measurement precision, and large data volumes, making traditional comparison techniques computationally infeasible and error-prone.
Innovation Solution
The implementation of a multi-stage process using a k-dimensional tree to perform event-level matching, followed by identifier-level matching, which normalizes data, builds sparse arrays, and applies thresholds to identify candidate matches and link panelist and subscriber identifiers, facilitating the fusion of demographic information and media exposure data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional comparison techniques are used to match identity across disparate data sets, then the matching process can be simple to implement, but the computational complexity becomes infeasible and error-prone due to large data volumes, asynchronous clocks, and differences in measurement precision
Solution Approach 1:
The patent segments the identity matching process into two distinct stages: event-level matching and identifier-level matching. Event-level matching uses k-dimensional trees to efficiently find candidate matches based on temporal and content proximity, while identifier-level matching applies statistical thresholds to confirm true matches. This segmentation reduces computational complexity from comparing all identifier pairs to first filtering candidates by events, then verifying with thresholds.
Solution Approach 2:
The patent performs preliminary normalization of data from different measurement systems before matching, including synchronizing timestamps to a reference clock and standardizing measurement precision. This preliminary action eliminates the need for complex real-time adjustments during matching and reduces errors from asynchronous clocks and precision differences.
2Measurement precision
If data from multiple audience measurement systems is merged to improve measurement accuracy, then the demographic information and media exposure data become more comprehensive, but the difficulty of detecting and measuring increases due to asynchronous clocks and differences in measurement precision
Solution Approach 1:
The patent creates equipotentiality by normalizing all data to a common reference frame before matching. Specifically, it synchronizes timestamps from different measurement systems to a reference clock and standardizes measurement precision levels. This eliminates potential differences (asynchronous clocks, precision variations) that would otherwise make detection and measurement difficult.
Solution Approach 2:
The patent changes key parameters of the input data including timestamp formats, precision levels, and synchronization references. By transforming这些数据 to compatible parameter states, the system enables accurate matching across disparate measurement systems without losing the benefits of comprehensive demographic and exposure data.
3Reliability
If exhaustive comparison of all data pairs is performed to ensure accurate matching, then the reliability of matching results improves, but the loss of time increases significantly due to large data volumes
Solution Approach 1:
The patent applies partial action by performing exhaustive comparison only within filtered candidate sets rather than across all possible pairs. The k-dimensional tree identifies a small subset of candidate matches based on temporal and content proximity, and threshold-based verification applies exhaustive checking only to these candidates. This reduces processing time from O(n²) to approximately O(n log n) while maintaining high reliability through the two-stage verification process.
Data Source
AI summary
Methods, apparatus, systems and articles of manufacture to perform identity matching across audience measurement systems are disclosed. An example apparatus includes processor circuitry to execute computer readable instructions to identify a first one of a plurality of search spaces in a data structure based on a query event, the data structure based on audience measurement events corresponding to a first audience measurement system, the query event corresponding to a second audience measurement system, determine the query event and a first audience measurement event in the search space are a candidate match when a distance between the query event and the first audience measurement event satisfies a threshold, generate a profile for a first user identifier based on the audience measurement events associated with the first audience measurement system based on the candidate match, the profile including the second user identifier but not the first user identifier, and provide the profile to the second audience measurement system without the first user identifier.