Multi-Graph Attention Network for TCM Syndrome Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional Chinese medicine (TCM) syndrome classification methods rely heavily on subjective diagnosis and fail to accurately consider the complex, nonlinear relationships between symptoms and syndrome elements, leading to inaccurate classification results.
Innovation Solution
A multi-graph attention network is employed to construct symptom-symptom and symptom-syndrome element graphs, aggregating features through graph attention mechanisms and using a multi-layer perceptron for classification, integrating symptom and syndrome element information for more comprehensive syndrome classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional machine learning methods are used to classify TCM syndrome types, then the classification process is simple and fast, but the accuracy is low due to inability to capture complex nonlinear relationships between symptoms and syndrome elements
Solution Approach 1:
The patent segments the TCM syndrome classification problem into multiple sub-problems by constructing separate graphs for different relationships: a symptom-symptom graph for symptom relationships, a symptom-syndrome element graph for symptom-syndrome element relationships, and a syndrome element-syndrome element graph for syndrome element relationships. Each graph captures a specific aspect of the complex relationships, and the results are integrated through attention mechanisms to achieve comprehensive and accurate classification.
Solution Approach 2:
The patent transitions from traditional flat table data representation to a multi-dimensional graph structure. By representing TCM data as graphs with nodes and edges, it adds relational dimensionality to the data, enabling the model to capture nonlinear relationships between symptoms, syndrome elements, and syndrome types that are not present in traditional tabular formats.
2Measurement precision
If only symptom-symptom relationships are considered, then the classification model is simple, but it fails to accurately reflect the complex relationships between symptoms and syndrome elements
Solution Approach 1:
The patent merges multiple graph structures (symptom-symptom graph, symptom-syndrome element graph, and syndrome element-syndrome element graph) into a unified multi-graph attention mechanism. This combination allows the model to simultaneously capture and integrate various types of relationships between symptoms and syndrome elements, achieving comprehensive relationship extraction without overwhelming complexity.
Solution Approach 2:
The patent introduces syndrome elements as intermediary nodes between symptoms and syndrome types. These intermediary nodes facilitate the capture of complex relationships by serving as bridges that connect symptoms to syndrome types through multiple pathways, enabling more accurate relationship extraction while maintaining manageable model structure.
3Productivity
If subjective diagnostic methods are used, then the diagnostic process is fast and relies on personal experience, but the results are highly subjective and difficult to promote
Solution Approach 1:
The patent replaces the mechanical system of subjective human diagnosis with an automated AI-based system. By using graph neural networks and attention mechanisms to process TCM data, the system automatically extracts syndrome type information from symptom data, eliminating the need for human practitioners' subjective judgment while maintaining diagnostic accuracy and consistency.
Data Source
AI summary
The present disclosure provides a traditional Chinese medicine (TCM) syndrome classification method based on multi-graph attention, which comprehensively considers the contribution of symptoms and syndrome elements in syndrome classification by constructing a graph structure, integrates a symptom-symptom graph and a symptom-syndrome element graph into classification, uses a multi-graph attention network to aggregate the features of symptoms and syndrome elements, and finally realizes syndrome classification through a multi-layer perceptron. At the same time, extensive experiments are carried out on real data sets, the effectiveness of the multi-graph attention network is verified, more accurate classification is realized, and better classification results have been achieved.