EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Multi-Task Genetic Algorithm-Optimized Deep Neural Network for Workload and Machine Failure Prediction in Data Centers

Shimaa Ouf

Journal of Intelligent Decision Making and Information Science · 08.09.2026 · 3 (11s) · 1853-1874

DOI: 10.17762/jidmis.v3.3200 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Intelligent Decision Making and Information Science, online erschienen am 08.09.2026. Verfasst von Shimaa Ouf. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Data centers are the backbone of modern digital services. However, workload variations and machine failures are major challenges that affect system reliability, utilization of resources, service availability, and operational efficiency. In most of the studies, workload prediction and machine failure prediction are generally considered as independent tasks and thus they cannot provide integrated support for predictive maintenance and intelligent resource management. To address this problem, this paper introduces a multi-task Genetic Algorithm-Optimized Deep Neural Network (GA-DNN) framework that performs workload prediction and machine failure prediction concurrently using the Google Cluster Trace 2019 dataset. The main contributions of this work are the unification of both prediction tasks into a multi-task learning framework and the application of Genetic Algorithm for automated optimization of the DNN learning rate and number of hidden neurons. The efficiency of the proposed approach is examined in two experimental settings. In the first experiment a baseline DNN without Genetic Algorithm based hyper-parameter optimization is evaluated for workload prediction and machine failure prediction separately. In the second experiment, the proposed multi-task GA-DNN uses the Genetic Algorithm to find the optimized hyperparameters and performs both prediction tasks. The baseline DNN reached accuracies of 78.77% and 84.93% for the workload prediction and machine failure prediction, respectively. The GA-DNN presented achieved 86.53% accuracy, 76.63% precision, 98.17% recall and 86.07% F1-score for workload prediction and 99.71% accuracy, 98.82% precision, 99.86% recall and 99.34% F1-score for machine failure prediction, respectively. These results of comparison indicate that hyperparameter optimization with Genetic Algorithm improves the predictive performance significantly over the non-optimized DNN baseline. The proposed framework provides an integrated and optimized approach for intelligent resource management and predictive maintenance of modern data centers.

Zitieren

APA

Ouf, S. (2026). A Multi-Task Genetic Algorithm-Optimized Deep Neural Network for Workload and Machine Failure Prediction in Data Centers. Journal of Intelligent Decision Making and Information Science, 3(11s), 1853-1874. https://doi.org/10.17762/jidmis.v3.3200

Vancouver

Ouf S. A Multi-Task Genetic Algorithm-Optimized Deep Neural Network for Workload and Machine Failure Prediction in Data Centers. Journal of Intelligent Decision Making and Information Science. 2026;3(11s):1853-1874. doi:10.17762/jidmis.v3.3200

BibTeX
@article{Ouf2026,
  author = {Ouf, Shimaa},
  title = {A Multi-Task Genetic Algorithm-Optimized Deep Neural Network for Workload and Machine Failure Prediction in Data Centers},
  journal = {Journal of Intelligent Decision Making and Information Science},
  year = {2026},
  volume = {3},
  number = {11s},
  pages = {1853--1874},
  doi = {10.17762/jidmis.v3.3200}
}
Kostenlos starten →