Clustering Student Learning Types with SHAP: Toward Interpretable Data-Driven Class Formation
ITM Web of Conferences · 2026 · 89 · 03014
DOI: 10.1051/itmconf/20268903014 ↗Kurz gefasst
Fachartikel in der Zeitschrift ITM Web of Conferences, online erschienen am 14.08.2026. Verfasst von Haobo Xu. Die Arbeit nennt 10 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
This study used K-Means clustering on a large student learning behavior dataset to find natural groupings among learners. Five clear types emerged. They are Rapid Improvers, Struggling Decliners, Elite Stable Performers, Trial Quitters, and Diligent Strivers. These groups differ in several ways. To make sense of the clusters, SHAP was applied to show which features drive the groupings. Recent accuracy, total attempts, and consistency mattered most across all groups. Improvement trends, logged attempts, and skill variety helped separate individual clusters. Turning clustering results into clear, feature-based explanations helps teachers and administrators see past the “black box” of data-driven grouping. The findings point toward mixed-ability class setups. For example, Rapid Improvers could be paired with Struggling Decliners. Trial Quitters would need extra support. This kind of grouping could build fairer and more effective learning spaces. Several limits should be noted. The data merges records from two platforms, which may introduce bias. K-Means assumes round clusters, which may not match real student patterns. SHAP values also depend on how well the surrogate classifier learned the labels. Future work should test these five types in other school settings and track how students move from one type to another over time.
Zitieren
APA
Xu, H. (2026). Clustering Student Learning Types with SHAP: Toward Interpretable Data-Driven Class Formation. ITM Web of Conferences, 89, 03014. https://doi.org/10.1051/itmconf/20268903014
Vancouver
Xu H. Clustering Student Learning Types with SHAP: Toward Interpretable Data-Driven Class Formation. ITM Web of Conferences. 2026;89:03014. doi:10.1051/itmconf/20268903014
BibTeX
@article{Xu2026,
author = {Xu, Haobo},
title = {Clustering Student Learning Types with SHAP: Toward Interpretable Data-Driven Class Formation},
journal = {ITM Web of Conferences},
year = {2026},
volume = {89},
pages = {03014},
doi = {10.1051/itmconf/20268903014}
}
EUVIMED