An Information-Theoretic Evaluation of Granule-Based Association Rule Mining Results
Entropy · 14.11.2026 · 28 (11) · 1260
DOI: 10.3390/e28111260 ↗Kurz gefasst
Fachartikel in der Zeitschrift Entropy, online erschienen am 14.11.2026. Verfasst von Bin Liu. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Evaluating the quality of knowledge discovered by association rule mining remains an open problem. Existing interestingness measures assess individual rules in isolation rather than how well a rule set stands in for the original dataset. This paper evaluates a mined rule set by an information-theoretic approach, with the view that mining is a data representation process. Four criteria are used: the compression ratio (C1), the entropy difference between the original dataset and the covered subset (C2), the entropy-estimation error between the mined rule set and the covered subset (C3), and C4, the fraction of the dataset’s total correlation retained by the maximum-entropy distribution consistent with the mined rules. Five case studies and experiments on five benchmark datasets demonstrate that the criteria effectively distinguish the quality of granule-based association rules at different levels. In particular, augmenting the support of a granule with its diversity reduces the estimation error by 80.5–97.8% on every dataset, whereas the association mapping raises C4 on all five datasets when it adds constraints beyond the condition granules.
Zitieren
APA
Liu, B. (2026). An Information-Theoretic Evaluation of Granule-Based Association Rule Mining Results. Entropy, 28(11), 1260. https://doi.org/10.3390/e28111260
Vancouver
Liu B. An Information-Theoretic Evaluation of Granule-Based Association Rule Mining Results. Entropy. 2026;28(11):1260. doi:10.3390/e28111260
BibTeX
@article{Liu2026,
author = {Liu, Bin},
title = {An Information-Theoretic Evaluation of Granule-Based Association Rule Mining Results},
journal = {Entropy},
year = {2026},
volume = {28},
number = {11},
pages = {1260},
doi = {10.3390/e28111260}
}
EUVIMED