Research on personalized teaching modeling based on dynamic entropy control clustering: Innovation and application of AKM-DWEC framework

The growing demand for personalized education necessitates intelligent clustering methods capable of handling dynamic, high-dimensional educational data while preserving privacy. This study proposes an Adaptive K-means framework with Dynamic Weight Entropy Control (AKM-DWEC), which addresses three c...

Ausführliche Beschreibung

Gespeichert in:

Bibliographische Detailangaben
Veröffentlicht in:	Systems and soft computing Jg. 7; S. 200353
1. Verfasser:	Chen, Xihua
Format:	Journal Article
Sprache:	Englisch
Veröffentlicht:	Elsevier B.V 01.12.2025 Elsevier
Schlagworte:	Adaptive clustering algorithm Differential privacy preservation Dynamic weight entropy control Multimodal data governance Personalized learning model Multimodal data governance Differential privacy preservation Personalized learning model Adaptive clustering algorithm Dynamic weight entropy control
ISSN:	2772-9419, 2772-9419
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Beschreibung
Zusammenfassung:	The growing demand for personalized education necessitates intelligent clustering methods capable of handling dynamic, high-dimensional educational data while preserving privacy. This study proposes an Adaptive K-means framework with Dynamic Weight Entropy Control (AKM-DWEC), which addresses three critical challenges in educational data analysis: (1) Dynamic feature relevance through entropy-regularized weight optimization, explicitly constraining weight uniformity with a novel regularization term λ∑i=1d(Wi−1d)2; (2) Real-time adaptation via incremental micro-cluster updating, reducing computational overhead by 85.3 %; and (3) privacy-utility balance achieved through dual perturbation combining Laplace noise (ε1=0.4) and Gaussian random projection (ε2=0.6). Experimental validation on 800 students' multimodal dataset demonstrates superior performance with 0.892 silhouette coefficient (25.1 % improvement over baseline K-means) and 91.3 % F1-score in learner categorization, while maintaining 91.2 % clustering purity under (ε=1.0, δ=10^5) - differential privacy. The framework's educational impact is evidenced by 28.7 % knowledge mastery improvement (p < 0.001) and 41.7 % reduction in recommendation errors. This work advances personalized education through its dynamic adaptation mechanism and rigorous privacy preservation, though cross-cultural applicability requires further investigation.
ISSN:	2772-9419 2772-9419
DOI:	10.1016/j.sasc.2025.200353