Machine Learning-Driven Adaptive Routing in Wireless Sensor Networks
International Journal on Advanced Science, Engineering and Information Technology · 25.06.2026 · 16 (3) · 874-879
DOI: 10.18517/ijaseit.16.3.13812 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal on Advanced Science, Engineering and Information Technology, online erschienen am 25.06.2026. Verfasst von Zaineb Talib Al-Azez und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Wireless Sensor Networks (WSNs) underpin critical monitoring infrastructure in industrial automation, environmental surveillance, smart agriculture, and healthcare telemetry, and their energy efficiency and routing intelligence directly determine operational lifetime and data reliability. Traditional routing protocols make decisions based on static or randomized rules that are agnostic to the dynamic state of individual nodes, leading to premature energy depletion, unbalanced cluster-head loads, and suboptimal path selection. This paper investigates the application of machine learning (ML) to adaptive routing in WSNs, making three principal contributions. First, we formulate next-hop routing as a three-class supervised classification problem and benchmark four ML models: Decision Tree, Random Forest, Support Vector Machine, and a proposed Deep Neural Network (DNN), trained on an eight-dimensional feature vector comprising residual energy, distance metrics, neighborhood density, queue state, RSSI, hop count, and traffic load. Second, we address residual-energy prediction as a regression task, comparing a Linear Baseline, Gradient Boosting, and a DNN regressor to enable proactive route adaptation before energy exhaustion. Third, we show through network-lifetime simulation that ML-assisted routing, instantiated with energy-weighted cluster-head election guided by DNN predictions, delays the First Node Dead event by 166 % (round 38 to round 101) and extends the Last Node Dead round to 216. The SVM achieves the highest classification accuracy (92.00 %), while the DNN achieves 90.40 % with superior generalization. For energy prediction, Gradient Boosting achieves the lowest RMSE (0.0085 J), while the DNN regressor achieves a competitive 0.0183 J with lower complexity. Feature-importance analysis identifies residual energy and RSSI as the two most informative routing features.
Zitieren
APA
Al-Azez, Z. T., Sakran, A. A., Al-Maliki, M. H. A., Ali, S. O. A., & Shaharudin, S. M. (2026). Machine Learning-Driven Adaptive Routing in Wireless Sensor Networks. International Journal on Advanced Science, Engineering and Information Technology, 16(3), 874-879. https://doi.org/10.18517/ijaseit.16.3.13812
Vancouver
Al-Azez ZT, Sakran AA, Al-Maliki MHA, Ali SOA, Shaharudin SM. Machine Learning-Driven Adaptive Routing in Wireless Sensor Networks. International Journal on Advanced Science, Engineering and Information Technology. 2026;16(3):874-879. doi:10.18517/ijaseit.16.3.13812
BibTeX
@article{AlAzez2026,
author = {Al-Azez, Zaineb Talib and Sakran, Ammar AbdRaba and Al-Maliki, Manar Hasan Ali and Ali, Salam Omar Alo and Shaharudin, Shazlyn Milleana},
title = {Machine Learning-Driven Adaptive Routing in Wireless Sensor Networks},
journal = {International Journal on Advanced Science, Engineering and Information Technology},
year = {2026},
volume = {16},
number = {3},
pages = {874--879},
doi = {10.18517/ijaseit.16.3.13812}
}
EUVIMED