Automated Data Model Configuration Selection for Knowledge Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The current process of selecting a data model configuration for training machine learning predictive models on knowledge graphs is manual, time-consuming, and prone to errors, especially when dealing with large and ever-changing datasets, which hinders the optimization of predictive accuracy for inferring relationships between entities like genes and diseases.
Innovation Solution
A system that automates the selection of data model configurations by receiving multiple configurations, extracting data models from a knowledge graph, generating predictive models, scoring their performance based on benchmark datasets, and selecting the most suitable configuration for improved predictive performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual ad hoc process is used to define data model configuration, then flexibility in customization is achieved, but time consumption and error rate increase significantly
Solution Approach 1:
The system enables self-service by allowing the data model configuration to automatically define and extract its own subsets from the knowledge graph based on predefined criteria such as confidence scores and relationship types, eliminating the need for manual configuration while maintaining adaptability through programmable parameters
Solution Approach 2:
The invention applies parameter changes by transforming the manual configuration process into an automated one where data model configurations are defined by adjustable parameters including confidence score thresholds, relationship type selections, and subset size limits, allowing flexible customization without manual intervention
2Measurement precision
If multiple data model configurations are evaluated through manual process, then optimal configuration can be identified, but productivity and experimentation speed decrease
Solution Approach 1:
The system performs preliminary action by automatically generating multiple data model configurations and their corresponding subsets before the evaluation phase, preparing all necessary data structures and configurations in advance to enable rapid experimentation and accurate comparison of different configurations
Solution Approach 2:
The invention implements continuity of useful action by establishing an automated workflow where data model configurations are continuously evaluated, compared, and refined without manual intervention, maintaining a continuous cycle of experimentation that simultaneously improves predictive accuracy and increases productivity
3Reliability
If manual selection of data model configuration is performed, then careful evaluation is possible, but reliability and consistency of results deteriorate due to human error
Solution Approach 1:
The system applies feedback by automatically evaluating each data model configuration's performance based on predictive accuracy metrics and using this feedback to consistently select the optimal configuration, eliminating human error while maintaining operational simplicity through automated decision-making loops
Data Source
AI summary
Method(s), apparatus, and system(s) are provided for selecting a data model configuration for use in training predictive models comprise receiving two or more data model configurations, extracting a data model for each of the two or more data model configurations from a knowledge graph, generating a separate predictive model for each of the extracted data models, scoring the output of each separate predictive model based on a benchmark data set, and selecting at least one data model configuration of the two or more data model configurations based on the output scores.