Dynamic Event Distribution Reweighting for Identity Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current social behavior analytics approaches face challenges in real-time identity resolution, event deduplication, and historical fidelity, often resulting in low accuracy and efficiency due to periodic clearing of distributions and reliance on batch processing, which recreates the 'cold start' issue and fails to effectively manage anomalous user behavior and duplicate events.
Innovation Solution
A system and method for constructing a distribution of interrelated event features by receiving a stream of events, extracting features, constructing and analyzing these features, and dynamically reweighting them to scale, enabling real-time identity resolution and efficient event deduplication through a probabilistic approach that adapts to changing user behavior and reduces false positives/negatives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If periodic clearing of distributions is performed, then memory usage is controlled, but historical fidelity deteriorates and accuracy decreases
Solution Approach 1:
The patent implements dynamic reweighting of event distributions where older events gradually decrease in weight rather than being abruptly cleared. This dynamic approach maintains historical fidelity by preserving the influence of past events while allowing the system to adapt to new patterns, resolving the contradiction between memory control and reliability.
Solution Approach 2:
The system changes the parameter of event weight over time through continuous reweighting mechanisms. Instead of binary presence/absence (clearing), events transition through varying weight parameters, maintaining their contribution to identity resolution accuracy while managing computational resources efficiently.
2Device complexity
If batch processing is used, then computational complexity is reduced, but real-time identity resolution capability deteriorates
Solution Approach 1:
The patent performs preliminary actions by continuously maintaining updated event distributions and weights in real-time rather than waiting for batch processing. This preliminary preparation of data structures enables rapid identity resolution when needed, eliminating the delay inherent in batch processing while managing complexity through incremental updates.
Solution Approach 2:
The system maintains continuous useful action by constantly updating event distributions and reweighting events as they occur. This continuous maintenance of accurate event weights enables real-time identity resolution without the interruptions and delays of batch processing, while computational complexity is managed through efficient incremental update algorithms.
3Ease of operation
If event distributions are maintained without reweighting, then processing simplicity is preserved, but accuracy in resolving identity deteriorates due to cold start issues
Solution Approach 1:
The patent implements feedback mechanisms where the system continuously monitors event distributions and automatically reweights events based on their recency and relevance. This feedback loop maintains high identity resolution accuracy by adapting to changing patterns while preserving processing simplicity through automated, rule-based reweighting that doesn't require complex manual intervention.
4Loss of information
If duplicate events are not deduplicated, then data completeness is maintained, but processing efficiency deteriorates
Solution Approach 1:
The patent uses copying techniques where duplicate events are identified and referenced rather than fully processed. The system maintains references to unique event distributions and reweights them appropriately, avoiding redundant processing of identical events while preserving the complete event history for accurate identity resolution.
Data Source
AI summary
A system, method, and computer-readable medium are disclosed for performing a distribution of interrelated event features operation. The distribution of interrelated event features includes: receiving a stream of events, the stream of events comprising a plurality of events; extracting features from the plurality of events; constructing a distribution of the features from the plurality of events; analyzing the distribution of the features from the plurality of events; and, dynamically reweighting the distribution of the features to scale a number of events contained within the distribution.