Graph Path Tracing via Meta-Path Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In graph structured data, tracing paths is challenging due to multiple interactions between entities, where the correlation between incoming and outgoing interactions is unknown, as seen in supply chains and social networks, with existing methods either relying on information sharing or being unable to model complex product production and fake news propagation accurately.
Innovation Solution
A method, system, and software that trace flow in graph structured data through meta-paths by defining types and attributes, generating a heterogeneous graph, associating nodes based on similarity, sampling positive and negative meta-paths, projecting nodes into embedded vectors, and tracing paths with similarity greater than a preset threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional path tracing methods are used in graph structured data, then the correlation between incoming and outgoing interactions can be traced, but the methods cannot handle multiple interactions between entities where the correlation is unknown
Solution Approach 1:
The patent transforms the path tracing problem from a deterministic correlation-based approach to a probabilistic similarity-based approach. By changing the fundamental parameter from known correlation to unknown similarity, the system can handle multiple interactions between entities where the specific correlation is not available. The similarity calculation uses node attributes and graph structure to estimate the likelihood of flow transmission along meta-paths.
Solution Approach 2:
The patent introduces meta-paths as intermediary concepts that represent sequences of relationship types between node types. These meta-paths serve as mediators that capture the flow of materials, information, or other entities through the graph structure without requiring direct knowledge of specific correlations between individual interactions. The meta-path similarity acts as an intermediary metric that bridges the gap between unknown specific correlations and overall path tracing accuracy.
2Adaptability or versatility
If agent-based models with specified rules are used to model propagation, then the propagation of negative shocks can be modeled, but the models cannot handle companies that produce multiple types of products
Solution Approach 1:
The patent creates a universal path tracing framework that works across different graph types (supply chains, social networks, contact networks) and different flow types (physical flow, information flow, infection routes). The system uses generic concepts like nodes, edges, meta-paths, and similarity calculations that can be applied universally without requiring product-specific or interaction-type-specific models. This allows companies producing multiple product types to be modeled within a single unified framework.
Solution Approach 2:
The patent extracts the essential structure of flow propagation from specific domain details. By separating the general path tracing mechanism from domain-specific assumptions (such as one-company-one-product assumptions), the system can handle complex real-world scenarios where companies produce multiple products. The extraction of core graph structure and meta-path concepts allows the model to focus on the flow pattern rather than specific product or company characteristics.
3Productivity
If related art methods relying on information sharing are used, then paths can be traced in some cases, but the methods are not able to be commonly applied
Solution Approach 1:
The patent enables the system to trace paths using only the graph structure and node attributes that are already available in the data. The similarity calculation between meta-paths is performed autonomously based on the inherent properties of the graph data, without requiring external information sharing or additional data collection. This self-service approach allows the method to be commonly applied across different scenarios where information sharing may not be feasible.
Data Source
AI summary
Trace flow of nodes in a graph structure dataset through meta-paths. For an input of a graph structure dataset, systems and methods can involve defining types and attributes for the plurality of nodes and the plurality of edges to generate a heterogenous graph from the input graph structure dataset; associating the plurality of nodes with each other based on similarity; defining a scheme for a meta-path between the plurality of nodes and the plurality of edges based on defined flows; sampling positive meta-paths and negative meta-paths for a selected node from the heterogenous graph based on the defined scheme; projecting the associated plurality of nodes into embedded vectors; and tracing paths of associated nodes having a similarity with the starting node specified by user greater than a preset threshold.