Graph Neural Network Competitor Detection for Sparse Supply Chain Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current competitor detection models in the financial services industry are unsuitable due to complexity, inability to account for directed and undirected relationships, and minimally annotated competitor relationships, leading to erroneous results and scalability issues.
Innovation Solution
A novel model using graph neural networks applies first-order and second-order proximity to detect competitors, leveraging a directed graph convolutional network (GCN) and Laplacian sharpening techniques to reconstruct node features, enhancing competitor detection and transactional relationship identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If current competitor detection models are used, then the system can identify competitors, but the models are too complex and cannot account for directed and undirected relationships
Solution Approach 1:
The patent segments the competitor detection problem into multiple sub-problems by separating directed and undirected relationships into different graph components. The directed graph captures supply chain relationships while the undirected graph captures competitor relationships, allowing each sub-graph to be processed with appropriate methods and reducing overall model complexity while maintaining detection accuracy.
Solution Approach 2:
The patent introduces an intermediary representation layer using graph embeddings that unify the processing of directed and undirected relationships. By converting both types of relationships into a common embedding space, the model can handle relationship complexity without requiring separate processing paths, thus reducing overall system complexity.
2Reliability
If manual data sifting methods are used, then the system can identify competitors, but the process is not scalable and produces erroneous results
Solution Approach 1:
The patent replaces manual data sifting with automated graph neural network processing. Instead of manually analyzing company data points, the system automatically processes large datasets through graph embeddings and neural network inference, enabling scalable competitor detection across the entire financial services industry without human intervention.
Solution Approach 2:
The patent transforms the detection approach by changing from manual analysis parameters to automated graph-based parameters such as node embeddings, edge weights, and graph structure features. This parameter transformation enables the system to process vast amounts of data automatically while maintaining high detection accuracy through mathematical optimizations.
3Adaptability or versatility
If current models are used, then the system can process data, but they cannot account for minimally annotated competitor relationships
Solution Approach 1:
The patent applies partial action by focusing the graph neural network on the most critical relationship attributes rather than requiring complete annotation of all relationship details. The model successfully detects competitor relationships even with minimal annotations by identifying patterns in the graph structure and node embeddings, thus reducing the data annotation requirement while maintaining detection effectiveness.
Data Source
AI summary
In some aspects, the techniques described herein relate to a method including: providing a graph neural network; and configuring the graph neural network to predict a competitor, the method comprising: receiving at least one dataset of nodes of supply chain companies, competitor companies, and customers and associated attached nodes' attributes; applying a first-order proximity to denote a local connection structure of some supply chain companies, competitor companies, and customers; applying a Laplacian Eigenmap to the first-order proximity to identify at least two positive pairs and at least two negative pairs; applying a pairwise ranking loss function that reduces the distance between the at least two positive pairs and increasing the distance between the at least two negative pairs; and based on an input identification of one company, ranking competitor companies of the company based on their Euclidean distances in the graph.