Predictive Entity Resolution Using Machine Learning Identity Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing entity resolution methods in computer network environments, such as those used in cybersecurity, rely on pre-defined rules that are labor-intensive and difficult to update dynamically, especially when definitive data is not available, making it challenging to accurately associate entities with attributes over time.
Innovation Solution
A predictive entity resolution system that uses machine learning techniques to generate identity models from evidence events, allowing for the probabilistic association of entities with attributes at specific times, enabling dynamic adaptation to changing behaviors and handling scenarios without definitive data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pre-defined rules are used for entity resolution, then entity association can be determined when definitive data is available, but the approach becomes labor-intensive and difficult to update dynamically
Solution Approach 1:
The patent replaces manual rule-based systems with machine learning models that automatically learn entity associations from data. The ML models substitute the mechanical process of creating and maintaining predefined rules, enabling dynamic adaptation without manual intervention while maintaining or improving association accuracy.
Solution Approach 2:
The system enables self-service by allowing the entity resolution model to automatically learn and update associations from available data sources without requiring manual rule preparation. The model serves itself by continuously improving through learning from evidence events, eliminating the need for labor-intensive rule maintenance.
2Reliability
If pre-defined rules are used for entity resolution, then definitive associations can be made when data is available, but the system cannot adapt dynamically to changing behaviors
Solution Approach 1:
The patent implements dynamics by using machine learning models that continuously learn from new evidence events and adapt to changing entity behaviors. The model transitions from static predefined rules to a dynamic system that updates its understanding of entity associations over time, allowing it to adapt to new patterns while maintaining reliable associations based on learned evidence.
3Productivity
If rule-based approaches are used, then labor-intensive rule preparation can be avoided when definitive data exists, but the approach fails when definitive data is not available
Solution Approach 1:
The patent applies universality by creating a machine learning-based entity resolution system that handles multiple scenarios: it can work with definitive data like traditional rules, but also effectively handles cases with incomplete or probabilistic data. The single ML-based approach replaces the need for different methods for different data types, improving both efficiency and reliability across all scenarios.
Data Source
AI summary
Predictive entity resolution uses a set of identity models to resolve an attribute to one or more associated entities, possibly with respective probabilities, at a particular time. The predictive entity resolution generates the sets of identity models from evidence events received from evidence sources. Each evidence event has at least one attribute and an associated time stamp.