Cross-Network Entity Profile Matching via Graph Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to effectively identify and associate profiles across vast networks, such as social media platforms, due to differences in information formats and the presence of duplicate names, making it challenging to link entities across different networks.
Innovation Solution
The method involves forming graphs of networked entities, using attributes like names, images, and interactions to predict whether profiles in one network are associated with the same entity in another network, by calculating probabilities and employing techniques like face recognition and decision trees to narrow down candidate nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If profile matching is performed across vast networks with millions of users, then the ability to identify entities across networks is improved, but the complexity and computational resources required increase significantly
Solution Approach 1:
The patent segments the profile matching process into multiple stages: first filtering candidates using unique identifiers (email, phone), then using graph-based methods to analyze interaction patterns, and finally applying machine learning classifiers. This multi-stage segmentation reduces the problem space at each step, making the overall system manageable despite the vast network sizes.
Solution Approach 2:
The patent introduces graph representations as an intermediary structure that captures interaction patterns between entities. These graphs serve as mediators that transform raw profile data into structured representations that can be more efficiently analyzed by matching algorithms, reducing the direct computational burden of comparing profiles across millions of users.
2Ease of manufacture
If traditional matching methods are used without unique identifiers, then the system is simpler to implement, but the precision of profile matching deteriorates due to duplicate names
Solution Approach 1:
The patent performs preliminary filtering using unique identifiers (email addresses, phone numbers) before applying more complex graph-based matching methods. This preliminary action eliminates many false candidates early in the process, allowing the system to maintain high precision without requiring complex implementations for all matching cases.
Solution Approach 2:
The patent changes the parameters used for matching from simple name-based criteria to multi-dimensional parameters including unique identifiers, graph interaction patterns, and machine learning scores. This parameter transformation enables precise matching even when names are duplicated, as the system evaluates multiple distinguishing features simultaneously.
3Measurement precision
If graph-based methods with multiple attributes are used for matching, then the accuracy of entity association is improved, but the computational time and resources required increase
Solution Approach 1:
The patent segments the graph analysis into hierarchical levels: first analyzing direct interactions, then expanding to second-degree connections, and only applying full machine learning classification to shortlisted candidates. This segmentation allows the system to leverage graph-based accuracy while limiting the computational time required by applying complex methods only where necessary.
Solution Approach 2:
The patent applies graph-based analysis and machine learning to a partial set of candidates rather than all possible profile pairs. By using unique identifiers to pre-filter the candidate set, the system performs excessive analysis (graph methods + ML) on fewer candidates, achieving high accuracy without the prohibitive computational cost of applying the same methods universally.
Data Source
AI summary
Described herein are systems and methods for predicting whether an entity associated with a profile in one network is the same entity that is associated with a profile in a second network, which networks may represent networks from different network services or may represent networks from the same network service. In embodiments, network graph features, including nodes and connections, may be used to predict a probability that the profiles in the two networks should be matched. In embodiments, additional or different factors may be included in the predicted probability, such as homophily, match probabilities of seed nodes, match probabilities of attribute-matched nodes, attribute-attribute co-occurrence probabilities, and the like.