EUVIMEDEUVIMED entdecken
Crossref

Designing Smart Learning Environments

MD Shahedul Islam Khan, Mowma Mazumder

US Journal of New Insights in Tech & Education · 25.08.2024

DOI: 10.69013/dq6667jh ↗

Kurz gefasst

Fachartikel in der Zeitschrift US Journal of New Insights in Tech & Education, online erschienen am 25.08.2024. Verfasst von MD Shahedul Islam Khan und 1 weiteren Autor:innen. Die Arbeit nennt 21 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Smart Learning Environments (SLEs) that combine the Internet of Things (IoT) with real-time sensor data are transforming how instruction is delivered, personalized, and evaluated. This study designs and empirically evaluates a sensor-driven SLE architecture that captures multimodal classroom data — attentiveness, movement, ambient conditions (light, noise, temperature, CO2), wearable heart-rate variability, and device-interaction logs — and applies machine learning (ML) techniques to model engagement and predict learning performance. Data were collected from a 12-week deployment across smart-enabled classrooms (N = 186 undergraduate students) and analyzed using six ML algorithms: logistic regression, decision tree, random forest, support vector machine, XGBoost, and a long short-term memory (LSTM) recurrent network. The LSTM model achieved the strongest predictive performance (accuracy = 0.913, F1-score = 0.902), followed closely by XGBoost (accuracy = 0.891); pairwise McNemar's tests confirmed that LSTM's advantage over all models except XGBoost was statistically significant. Random-forest feature-importance analysis identified time-on-task, device-interaction frequency, and wearable heart-rate variability as the strongest predictors of learning outcomes, an ablation analysis confirmed time-on-task as the single most consequential modality, and K-means clustering revealed three distinct and dynamically shifting engagement profiles among learners. Correlation and ROC-AUC analyses further showed that behavioral and physiological indicators discriminated at-risk learners more effectively than static environmental measures. The findings demonstrate that IoT-enabled sensor fusion, combined with ML-based analytics, can support real-time, personalized instructional adjustments and early identification of disengaged learners. The study concludes with design recommendations for scalable, privacy-conscious SLE architectures and directions for future research on adaptive, sensor-informed pedagogy.

Zitieren

APA

Khan, M. S. I., & Mazumder, M. (2024). Designing Smart Learning Environments. US Journal of New Insights in Tech & Education. https://doi.org/10.69013/dq6667jh

Vancouver

Khan MSI, Mazumder M. Designing Smart Learning Environments. US Journal of New Insights in Tech & Education. 2024. doi:10.69013/dq6667jh

BibTeX
@article{Khan2024,
  author = {Khan, MD Shahedul Islam and Mazumder, Mowma},
  title = {Designing Smart Learning Environments},
  journal = {US Journal of New Insights in Tech & Education},
  year = {2024},
  doi = {10.69013/dq6667jh}
}
Kostenlos starten →