A Comparative Analysis of Feature Selection Methods for Imbalanced Data Classification using PSO-Optimized Machine Learning Models
JOIV : International Journal on Informatics Visualization · 31.05.2026 · 10 (3) · 1090
DOI: 10.62527/joiv.10.3.4374 ↗Kurz gefasst
Fachartikel in der Zeitschrift JOIV : International Journal on Informatics Visualization, online erschienen am 31.05.2026. Verfasst von Willy Riyadi und 5 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).
Abstract (Originalsprache)
The rapid proliferation of the Internet of Medical Things (IoMT) has revolutionized healthcare delivery but has simultaneously expanded the attack surface, introducing critical security vulnerabilities. Developing practical Intrusion Detection Systems (IDS) for these environments is complicated by high-dimensional traffic data, severe class imbalance, and the requirement for low-latency detection. This study proposes a comprehensive machine learning framework designed to enhance the robustness of intrusion detection in IoMT networks. The primary contribution of this work is a systematic benchmarking of feature selection paradigms—Embedded (Elastic Net), Filter (Symmetrical Uncertainty), and Wrapper (Recursive Feature Elimination)—integrated within a Particle Swarm Optimization (PSO) framework. Using the massive CIC-IoMT2024 dataset, we implement a scientifically justified two-stage imbalance-handling strategy (Stratified Downsampling followed by SMOTEENN) to enable computationally feasible training on millions of records. Experimental results demonstrate that the wrapper-based RFE method yields the highest stability. The XGBoost-RFE configuration emerged as the superior model, achieving a Test F1-Score of 99.49% and a Mean CV F1-Score of 0.9970. Statistical validation via the Wilcoxon signed-rank test confirms that RFE significantly outperforms filter and embedded methods ( < 0.05). Furthermore, inference latency analysis confirms the framework's suitability for real-time deployment (approx. 12.2 µs/sample), validating its suitability for real-time detection.
Zitieren
APA
Riyadi, W., Kurniabudi, Jasmir, Kisbianty, D., Novianto, Y., & Sika, X. (2026). A Comparative Analysis of Feature Selection Methods for Imbalanced Data Classification using PSO-Optimized Machine Learning Models. JOIV : International Journal on Informatics Visualization, 10(3), 1090. https://doi.org/10.62527/joiv.10.3.4374
Vancouver
Riyadi W, Kurniabudi, Jasmir, Kisbianty D, Novianto Y, Sika X. A Comparative Analysis of Feature Selection Methods for Imbalanced Data Classification using PSO-Optimized Machine Learning Models. JOIV : International Journal on Informatics Visualization. 2026;10(3):1090. doi:10.62527/joiv.10.3.4374
BibTeX
@article{Riyadi2026,
author = {Riyadi, Willy and Kurniabudi, - and Jasmir, - and Kisbianty, Desi and Novianto, Yudi and Sika, Xaverius},
title = {A Comparative Analysis of Feature Selection Methods for Imbalanced Data Classification using PSO-Optimized Machine Learning Models},
journal = {JOIV : International Journal on Informatics Visualization},
year = {2026},
volume = {10},
number = {3},
pages = {1090},
doi = {10.62527/joiv.10.3.4374}
}
EUVIMED