An Explainable Ensemble Machine Learning Framework for Sustainable IoT-based Predictive Maintenance of University Computer Systems
Computer and Decision Making: An International Journal · 13.11.2026 · 4 (1) · 291-307
DOI: 10.1922/comdem.v4i1.19132 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 13.11.2026. Verfasst von Analene Montesines Nagayo und 5 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The paper presents the design architecture of an Internet of Things (IoT)- and Artificial Intelligence (AI)-driven equipment monitoring system for university computer laboratories that enables sustainable predictive maintenance by facilitating early fault detection, extending the equipment service lifespan, reducing electronic waste, and minimizing maintenance costs. The proposed system integrates IoT-based sensing, edge computing using a Raspberry Pi 5 (RPi5), cloud monitoring, and an ensemble machine learning model (EML) to predict and classify the health risk status of laboratory computers. Key operating parameters monitored include AC operating voltage, temperature, humidity, PM2.5 dust levels, computer age, failure history, computer system utilization and error logs. These inputs are utilized for predicting and classifying equipment health risk status into three categories: Healthy, Warning, or Critical. To improve prediction performance and robustness, the proposed system employs a majority-voting ensemble model that combines four machine learning (ML) classifiers, namely Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting (GB) and Naïve Bayes (NB). An Explainable Artificial Intelligence (XAI) module is also incorporated to give transparent interpretation of each prediction, which assists computer laboratory technicians in better understanding the model outputs and make well-informed maintenance decisions. Based on the hold-out test set evaluation, the EML model achieved an accuracy of 93.30%, an F1-score of 92.35%, a precision of 94.24%, and a recall of 90.85%. Furthermore, experimental validation performed through real-time deployments in actual university computer laboratories demonstrated that the proposed EML model attained an accuracy of 96.25%. The predicted equipment health risk classifications strongly agreed with the assessments of expert technicians and established computer system operating guidelines.
Zitieren
APA
Nagayo, A. M., Vikraman, B. P., Mahadevan, V., Guduri, N. V. R., Gadi, R. B., & Cabas, F. A. (2026). An Explainable Ensemble Machine Learning Framework for Sustainable IoT-based Predictive Maintenance of University Computer Systems. Computer and Decision Making: An International Journal, 4(1), 291-307. https://doi.org/10.1922/comdem.v4i1.19132
Vancouver
Nagayo AM, Vikraman BP, Mahadevan V, Guduri NVR, Gadi RB, Cabas FA. An Explainable Ensemble Machine Learning Framework for Sustainable IoT-based Predictive Maintenance of University Computer Systems. Computer and Decision Making: An International Journal. 2026;4(1):291-307. doi:10.1922/comdem.v4i1.19132
BibTeX
@article{Nagayo2026,
author = {Nagayo, Analene Montesines and Vikraman, Bindu Puthentharayil and Mahadevan, Vanitha and Guduri, Naga Venkata Ramakrishna and Gadi, Rahila Begum and Cabas, Fred Abogadie},
title = {An Explainable Ensemble Machine Learning Framework for Sustainable IoT-based Predictive Maintenance of University Computer Systems},
journal = {Computer and Decision Making: An International Journal},
year = {2026},
volume = {4},
number = {1},
pages = {291--307},
doi = {10.1922/comdem.v4i1.19132}
}
EUVIMED