Binary Search Algorithm-Inspired Hybrid Feature Selection
7th International Conference on Research in Computational Intelligence and Communication Networks, ICRCICN 2025, Hybrid, Kalyani, Hindistan, 20 - 21 Aralık 2025, ss.623-628, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/icrcicn68210.2025.11364867
- Basıldığı Şehir: Hybrid, Kalyani
- Basıldığı Ülke: Hindistan
- Sayfa Sayıları: ss.623-628
- Anahtar Kelimeler: Aegean Wi-Fi Intrusion Dataset, binary search algorithm, hybrid feature selection, Wi-Fi intrusion detection, wrapper feature selection
- Kocaeli Üniversitesi Adresli: Evet
Özet
High-dimensional data is a critical element that significantly impacts the detection accuracy and computational efficiency of network intrusion detection models. Models in this field rely heavily on the selection of an optimal subset. Although wrapper-based feature selection methods are effective, they often require considerable computational time. In this study, a hybrid feature selection method, inspired by the Binary Search Algorithm, is proposed. Aegean Wi-Fi Intrusion Dataset 3 is adopted for the study. Feature importance extraction is conducted using tree-based ensemble models, namely Extra Trees, Gradient Boosting Trees (GBT), and Random Forest algorithms. Subsequently, subsets are selected and tested with K-Nearest Neighbors (KNN), LightGBM, and XGBoost classifiers. The results show that the GBT+KNN model is the best-performing, with both accuracy and an F1-score of 0.9982. For future work, it is planned to refine the proposed method by integrating optimization algorithms to further improve feature selection efficiency and detection accuracy.