Graph-Supplemented Disease Prediction Models for Sparse Biometric Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing disease prediction systems face challenges in achieving high prediction accuracy due to variations in human health and vital data, leading to deficiencies in training data that are not uniform across individuals.
Innovation Solution
A learning model generation apparatus and method that generates a graph from biometric information, supplements deficiencies in the graph, and uses the supplemented data for machine learning to create a prediction model, improving accuracy by integrating nodes and edges representing relationships between data points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning is performed using raw biometric data as training data, then the system can be simple and fast to implement, but prediction accuracy deteriorates due to data deficiencies and non-uniformity across individuals
Solution Approach 1:
The patent applies preliminary action by constructing a graph structure from biometric data before machine learning. The graph generation unit creates nodes representing data points and edges representing relationships between them, organizing the data in advance. This preprocessing step prepares the data in a structured format that facilitates more accurate prediction while managing complexity through systematic organization rather than raw data processing.
Solution Approach 2:
The patent transforms the data from a traditional tabular format into a graph structure, adding a dimensional aspect that represents relationships between data points. By converting biometric data into nodes and edges with associated features, the system captures contextual relationships that are lost in conventional flat data structures, thereby improving prediction accuracy without proportionally increasing processing complexity.
2Measurement precision
If more training data is collected to improve prediction accuracy, then prediction performance improves, but data non-uniformity and deficiencies increase because different individuals have different types of data
Solution Approach 1:
The patent applies universality by creating a graph structure that can accommodate various types of biometric data from different individuals in a unified framework. The graph nodes and edges can represent different data types (health data, vital data, lifestyle data) and their relationships, allowing the system to process non-uniform data from multiple sources consistently. This universal structure enables the model to learn from diverse data while maintaining adaptability to individual variations.
Solution Approach 2:
The patent changes the parameters of data representation by transforming raw biometric values into graph-based features. Instead of using original data values directly, the system extracts features from graph structures (such as node properties, edge relationships, and graph-level statistics) that capture essential patterns while being invariant to individual data variations. This parameter transformation allows the model to focus on meaningful patterns rather than raw value differences.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A learning model generation apparatus 10 comprises: a graph generation unit 11 which generates, from a data group including biometric information of persons and information indicating the presence or absence of occurrence of diseases in the persons, a graph composed of nodes representing data points and edges representing relationships between the nodes; a graph supplementation unit 12 which supplements the generated graph for a deficiency therein; and a model generation unit 13 which generates, from the supplemented graph, a data group in which the deficiency is supplemented, performs machine learning using the generated data group as training data, and generates a prediction model for predicting the occurrence of diseases in a person.