Real-Time Intrusion Detection System Using PCA and Ensemble Machine Learning
I. A. Adeyemo, O. G. Olatoye, D. G. Amusan, and B. A. Abdulsalami
We compared Random Forest and XGBoost on network traffic after reducing feature dimensions with PCA and addressing class imbalance with SMOTE. Looking at errors by class showed that a model with high overall accuracy could still miss attacks and scans.