Clustered interpretability - Developing metrics and a framework for explainable machine learning in clustered datasets

Aug 3, 2026

Cluster-heterogenous data—where feature distributions differ systematically across subgroups—frequently arises in machine learning research and complicates analysis. This work addresses a gap in interpretable machine learning by extending explainability beyond global and local perspectives to the cluster level, where feature-outcome relationships must be understood for groups. It introduces novel metrics to quantify interpretability of cluster solutions based on heterogeneous feature-outcome relationships, then proposes a metaheuristic partition-fitting algorithm to optimize them with within- and between-cluster interpretability. Experiments on two real-world healthcare datasets show equal or improved predictive performance versus conventional clustering while producing more interpretable cross-cluster feature-outcome relationships, supporting explainable AI decision making in information systems.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages2

Reading Time5 min

AccessFree to download

Abstract

Problem: cluster-heterogenous data and the interpretability gap

Proposed contribution: metrics for cluster-solution interpretability

Optimization approach: metaheuristic fitting of cluster partitions

Evaluation: healthcare datasets and comparison to standard clustering

Association for Information Systems AIS Electronic Library (AISeL) | AMCIS 2024 TREOs | AIS TREO Papers | | --- | --- | | 8-16-2024
Clustered interpretability: Developing metrics and a framework for explainable machine learning in clustered datasets
Matthew Baucum
Colorado State University, [baucum@usc.edu](baucum@usc.edu)
Follow this and additional works at: [https://aisel.aisnet.org/treos_amcis2024](https://aisel.aisnet.org/treos_amcis2024) | | Recommended Citation Baucum, Matthew, "Clustered interpretability: Developing metrics and a framework for explainable machine learning in clustered datasets" (2024) . AMCIS 2024 TREOs. 83. [https://aisel.aisnet.org/treos_amcis2024/83](https://aisel.aisnet.org/treos_amcis2024/83) This material is brought to you by the AIS TREO Papers at AIS Electronic Library (AISeL) . It has been accepted for inclusion in AMCIS 2024 TREOs by an authorized administrator of AIS Electronic Library (AISeL) . For more information, please [contact elibrary@aisnet.org](contact elibrary@aisnet.org). Clustered interpretability: Developing metrics and a framework for explainable machine learning in clustered datasets TREO Talk Paper Matt Baucum Colorado State University [mbaucum1@alum.utk.edu](mbaucum1@alum.utk.edu) Babak Aslani George Mason University [baslani@gmu.edu](baslani@gmu.edu) Meysam Rabiee University of Colorado Denver [Meysam.rabiee@ucdenver.edu](Meysam.rabiee@ucdenver.edu) Abstract Cluster-heterogenous data – i.e., datasets in which feature distributions systematically differ across subgroups – is commonly encountered in machine learning research. Because cluster-based heterogeneity complicates the data analysis process, it is especially important that machine learning models trained on clustered datasets follow best practices for interpretability and explainability. Yet there is currently a literature gap regarding how the principles of interpretable machine learning (IML) should be applied to models trained on clustered datasets. Most IML research focuses on interpretability at either the global level (i.e., understanding feature-outcome relationships across the entire dataset) or the local level (understanding feature-outcome relationships for individual training instances), with no well-established techniques for understanding feature-outcome relationships at the cluster level. Other research has focused on the interpretability of unsupervised clustering (i.e., clustering features based on their data values), but quantifying and optimizing the interpretability of clusters based on their feature-outcome relationships isan open research question. In this work, we propose novel metrics for quantifying the interpretability of cluster solutions based on the clusters’heterogenous feature-outcome relationships. We then develop a metaheuristic algorithm for fitting cluster partitions that optimize these metrics. Our approach emphases both within-cluster interpretability (i.e., ensuring each cluster is well-described by a parsimonious set of feature-outcome relationships) and between-cluster interpretability (i.e., ensuring that clusters’ feature-outcome relationships differ in understandable ways) . Using two real-world healthcare datasets, we demonstrate that machine learning models trained on interpretable cluster solutions perform just as well (or better) than models trained on traditionally-fit clusters (e.g., k-means, DB-scan, etc.), and that the heterogeneous feature-outcome relationships across clusters are more interpretable under our approach. We discuss our framework’s implications for the information systems field’s increasing emphasis on explainable AI-driven decision making (Bauer et al. 2023) . References Bauer, K., von Zahn, M., & Hinz, O. (2023).“Expl (AI) ned: The impact of explainable artificial intelligence on users’ information processing.” Information Systems Research, 34(4), 1582-1602.

FAQs

What interpretability gap does the work target for clustered datasets?

It targets the lack of well-established techniques to explain feature-outcome relationships at the cluster level, beyond existing global or local interpretability methods.

How does the approach quantify interpretability of cluster solutions?

It proposes novel metrics that evaluate interpretability based on the clusters’ heterogeneous feature-outcome relationships, capturing both within-cluster and between-cluster interpretability.

What results are reported when using the interpretable cluster framework on real-world data?

On two healthcare datasets, models trained with interpretable cluster solutions perform as well as or better than models trained on traditional clustering methods, with clusters whose heterogeneous feature-outcome relationships are more interpretable.

Aug 3, 2026

Upload by Theodore