Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Leveraging Knowlegde Graphs for Interpretable Feature Generation

Mohamed Bouadi, Arta Alavi, Salima Benbernou, Mourad Ouziri

TL;DR

This work tackles the need for interpretable feature engineering by automating feature generation with a knowledge-graph-guided framework, KRAFT. It combines a DRL-based Generator with a DL-based Knowledge-based Discriminator to produce features that maximize predictive performance while ensuring interpretability through KG semantics and SWRL rules. Empirical results across diverse datasets show competitive accuracy gains and enhanced interpretability, with feature importance analyses and SHAP confirming domain-aligned, readable features. The approach offers a model-agnostic, scalable pathway to integrate domain knowledge into AutoFE for practical, critical-context applications.

Abstract

The quality of Machine Learning (ML) models strongly depends on the input data, as such Feature Engineering (FE) is often required in ML. In addition, with the proliferation of ML-powered systems, especially in critical contexts, the need for interpretability and explainability becomes increasingly important. Since manual FE is time-consuming and requires case specific knowledge, we propose KRAFT, an AutoFE framework that leverages a knowledge graph to guide the generation of interpretable features. Our hybrid AI approach combines a neural generator to transform raw features through a series of transformations and a knowledge-based reasoner to evaluate features interpretability using Description Logics (DL). The generator is trained through Deep Reinforcement Learning (DRL) to maximize the prediction accuracy and the interpretability of the generated features. Extensive experiments on real datasets demonstrate that KRAFT significantly improves accuracy while ensuring a high level of interpretability.

Leveraging Knowlegde Graphs for Interpretable Feature Generation

TL;DR

This work tackles the need for interpretable feature engineering by automating feature generation with a knowledge-graph-guided framework, KRAFT. It combines a DRL-based Generator with a DL-based Knowledge-based Discriminator to produce features that maximize predictive performance while ensuring interpretability through KG semantics and SWRL rules. Empirical results across diverse datasets show competitive accuracy gains and enhanced interpretability, with feature importance analyses and SHAP confirming domain-aligned, readable features. The approach offers a model-agnostic, scalable pathway to integrate domain knowledge into AutoFE for practical, critical-context applications.

Abstract

The quality of Machine Learning (ML) models strongly depends on the input data, as such Feature Engineering (FE) is often required in ML. In addition, with the proliferation of ML-powered systems, especially in critical contexts, the need for interpretability and explainability becomes increasingly important. Since manual FE is time-consuming and requires case specific knowledge, we propose KRAFT, an AutoFE framework that leverages a knowledge graph to guide the generation of interpretable features. Our hybrid AI approach combines a neural generator to transform raw features through a series of transformations and a knowledge-based reasoner to evaluate features interpretability using Description Logics (DL). The generator is trained through Deep Reinforcement Learning (DRL) to maximize the prediction accuracy and the interpretability of the generated features. Extensive experiments on real datasets demonstrate that KRAFT significantly improves accuracy while ensuring a high level of interpretability.
Paper Structure (18 sections, 7 equations, 6 figures, 4 tables)

This paper contains 18 sections, 7 equations, 6 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Problem Definition
  3. Defining Interpretability
  4. Feature Engineering
  5. Proposed Approach
  6. An Overview of KRAFT
  7. Generator
  8. Semantic Vectorization
  9. Feature Generation
  10. Knowledge-based discriminator
  11. Knowledge representation
  12. Feature Selection
  13. Analysis and Interpretation
  14. Experiments
  15. Experimental Settings
  16. ...and 3 more sections

Figures (6)

  • Figure 1: Automated Machine Learning workflow.
  • Figure 2: An overview of KRAFT architecture.
  • Figure 3: A sample of the KG.
  • Figure 4: Effect of high-order features on accuracy improvement.
  • Figure 5: Comparing feature importance of raw features (in orange) and generated features with KRAFT (in blue).
  • ...and 1 more figures

Theorems & Definitions (3)

  • Example 1
  • Definition 1
  • Definition 2