Interactive Knowledge-Graph Feature Engineering for AutoML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current feature engineering in automatic machine learning (AutoML) lacks integration of domain knowledge, relying on static rules that do not consider the relationships and concepts inherent in the data, leading to suboptimal model performance.
Innovation Solution
A system and method that utilizes a knowledge graph to map features to concepts, traverse the graph to identify new features, and integrate these with existing features for machine learning model training, incorporating domain knowledge through an interactive user interface.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If static feature engineering rules are applied, then automation is achieved, but domain knowledge is not integrated
Solution Approach 1:
The patent introduces a knowledge graph as an intermediary structure that connects existing features to domain knowledge concepts. The system maps features to nodes in the knowledge graph, allowing automated feature engineering to leverage domain knowledge without manual intervention. This intermediary enables the bridge between automation and domain knowledge integration.
Solution Approach 2:
The system performs preliminary mapping of features to knowledge graph nodes before the actual feature engineering process. By pre-establishing these connections and identifying candidate features through graph traversal in advance, the system prepares domain knowledge context that guides subsequent automated feature generation, ensuring both automation and domain knowledge integration.
2Measurement precision
If domain knowledge is integrated through knowledge graphs, then feature accuracy improves, but system complexity increases
Solution Approach 1:
The patent segments the feature engineering process into distinct modules: feature mapping to knowledge graph nodes, graph traversal to find candidate features, candidate evaluation, and final feature selection. This segmentation allows each component to be independently optimized and managed, reducing overall system complexity while maintaining high feature accuracy through specialized processing at each stage.
Solution Approach 2:
The knowledge graph serves multiple functions: it stores domain knowledge, provides feature mappings, guides feature generation, and enables candidate evaluation. This multi-functionality reduces the need for separate systems for each task, thereby managing complexity while achieving high feature accuracy through a unified domain knowledge infrastructure.
3Loss of information
If manual feature selection is performed, then domain knowledge is utilized, but automation is reduced
Solution Approach 1:
The system enables self-service feature engineering by automatically traversing the knowledge graph, identifying candidate features based on domain knowledge relationships, and selecting optimal features without manual intervention. The automated system serves itself by leveraging the structured knowledge graph to perform tasks that traditionally required manual domain expert involvement, thus maintaining both automation and domain knowledge utilization.
Data Source
AI summary
A dataset including features and values associated with the features can be received. Each of the features in the dataset can be mapped to a corresponding node in a knowledge graph based on the concept represented by the corresponding node. The knowledge graph can be traversed to find a candidate node connected to at least one mapped node, the candidate node not being mapped to a feature in the dataset. A concept associated with the candidate node can be identified as a new feature. A machine learning model pipeline can use the features in the dataset and the new feature to select a subset of features for training a machine learning model.