Graph Embedding Model for Node Classification Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing graph-based classification systems independently process graph and non-graph data, leading to inaccurate and inefficient classification results due to the dissonance between numerical node representations and classification labels.
Innovation Solution
A method and apparatus for graph embedding that combines graph data and non-graph data using a neural network-based model, incorporating an accuracy parameter to generate numerical node representations by minimizing a combined objective function that accounts for user relationships and classification labels, thereby enhancing classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If graph data and non-graph data are processed independently in classification tasks, then the processing complexity is reduced and computational efficiency is improved, but classification accuracy deteriorates due to dissonance between numerical node representations and classification labels
Solution Approach 1:
The patent combines graph data processing and non-graph data processing into a unified neural network-based graph embedding model. The model integrates graph structure information (from graph data) and node attribute information (from non-graph data) to generate numerical node representations that are consistent with classification labels, thereby improving classification accuracy while maintaining computational efficiency through a single integrated processing framework.
2Device complexity
If only graph data is used to generate numerical node representations, then the model complexity is reduced and training speed is improved, but classification accuracy deteriorates due to lack of non-graph data utilization
Solution Approach 1:
The patent merges graph data and non-graph data into a unified neural network model that processes both data types simultaneously. The model architecture integrates graph convolutional networks for processing graph data with dense layers for processing non-graph data, combining their outputs to generate comprehensive numerical node representations that improve classification accuracy without excessive complexity increase.
Solution Approach 2:
The neural network-based graph embedding model serves multiple functions: it processes graph structure data, processes node attribute data, generates numerical node representations, and performs classification. This multi-functional approach allows the model to utilize both graph and non-graph data effectively within a single unified framework.
3Loss of time
If numerical node representations are generated without considering classification labels, then the graph embedding process is simplified and execution time is reduced, but classification accuracy deteriorates due to misalignment between node representations and classification categories
Solution Approach 1:
The patent performs preliminary action by training the neural network model with classification labels during the graph embedding process. The model learns to generate numerical node representations that are aligned with classification categories from the beginning, rather than performing separate alignment steps later. This preliminary integration of classification objectives into the embedding process improves accuracy without significant time penalty.
Solution Approach 2:
The patent implements feedback mechanisms where classification labels are used to guide the generation of numerical node representations. The model receives feedback from classification outcomes and adjusts its representation learning process accordingly, ensuring that node embeddings are optimized for both graph structure preservation and classification accuracy simultaneously.
Data Source
AI summary
System, method, device, and program for graph embedding based on graph data and non-graph data are provided. The method and processes may be executed by at least one processor and may include receiving graph data associated with one or more users, and receiving classification data associated with the one or more users, wherein the classification data comprises non-graph data associated with the one or more users. The method and processed may further include generating an accuracy parameter based on the classification data associated with the one or more users, wherein the accuracy parameter indicates an accuracy of neural network-based classification results based on the classification data; and generating numerical node representation using a neural network-based graph embedding model associated with the one or more users based on the graph data, the classification data, and the accuracy parameter.