Machine learning models reveal hidden feature interactions for better explanations
Beyond feature attribution: Recovering latent interaction structures in machine learning models via localized CBFI
Machine Learning
Summary
It's often hard to understand why complex computer programs give certain answers because they mix up the effects of different pieces of information. The authors developed a new way to break down these answers and clearly show how individual features and their combinations affect predictions. They tested their method on tasks like rating wine quality and predicting insurance costs, proving it can find hidden patterns that other methods miss. This approach helps people trust and check machine learning decisions more thoroughly.
What this means in practice
- •For medical data teams: Identify which combinations of patient data features jointly influence insurance cost predictions to better understand model reasoning.
- •For financial risk analysts: Detect complex feature interactions in credit scoring models to improve transparency and auditing of automated decisions.
Authors
Sejong Oh
Abstract
Background: Understanding individual predictions in complex tabular models remains challenging, as existing additive frameworks like SHAP and LIME often obscure structural logic by absorbing nonlinear interactions into individual feature scores. Exposing these latent interaction structures is essential for automated auditing in high-stakes environments where understanding the logic of model behavior is as essential as predictive accuracy. Methods: To address this limitation, we introduce the localized case-based feature importance (CBFI) framework. Unlike traditional numerical attribution, our method decomposes instance-specific predictions into structurally distinct components, effectively separating independent main effects (G1) from pure interactions (G4) to capture explicit feature synergies and provide a principled approach for diagnosing feature combinations. Results: Comprehensive experiments on wine quality classification and medical insurance regression demonstrate that localized CBFI successfully disentangles complex interactions hidden by additive models. Furthermore, fidelity tests and synthetic ground-truth benchmarks confirm that the identified interactions correspond precisely to the nonlinear structural dependency patterns embedded in complex architectures, including gradient boosting and deep neural networks, thereby successfully shifting the explainable AI paradigm from simple numerical attribution to the recovery of interaction-driven contributions.