Ontology-Enhanced Graph Embeddings for Medication Side Effect Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning models for estimating side effects of medications face accuracy issues due to reliance solely on similarities between feature vectors, which fail to account for patient attributes and disease relationships effectively.
Innovation Solution
A machine learning system that converts case data into graph data and merges ontology information to generate embedding vectors, allowing for the training of a model that appropriately reflects ontology information and improves estimation accuracy by using these vectors as training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models rely solely on similarities between feature vectors, then the model structure remains simple, but estimation accuracy deteriorates due to failure to account for patient attributes and disease relationships
Solution Approach 1:
The patent merges ontology information with patient case data to create integrated training datasets. The ontology embeddings capture disease relationships and medication interactions, which are combined with patient-specific features to form comprehensive input vectors for the machine learning model, thereby improving estimation accuracy while maintaining manageable model complexity
Solution Approach 2:
The patent performs preliminary processing by converting ontology data into embedding vectors before training the machine learning model. This pre-computation of ontology representations allows the model to efficiently utilize structured knowledge about disease relationships and medication interactions without increasing the complexity of the model architecture itself
2Measurement precision
If ontology information is integrated into feature vectors, then estimation accuracy improves, but data processing complexity increases
Solution Approach 1:
The patent introduces embedding vectors as an intermediary representation of ontology information. These embeddings serve as a bridge between the structured ontology data and the machine learning model, transforming complex relational knowledge into compact numerical vectors that are easy to process and integrate with patient features
Solution Approach 2:
The patent transforms ontology data from its original structured format into embedding vector representations, changing the parameter space from discrete categorical relationships to continuous numerical vectors. This parameter transformation simplifies the integration process and enables efficient processing by standard machine learning algorithms
Data Source
AI summary
A non-transitory computer-readable storage medium storing an estimation program that causes at least one computer to execute a process, the process includes inputting training data that includes a vector of graph data, a vector of ontology, and a label; training a machine learning model based on a loss function acquired by the label and a value obtained by merging a value of an activation function acquired with the vector of the graph data and a value of the activation function acquired with the vector of the ontology.