Improving Network Classification Accuracy through Feature Clustering and Ensemble Machine Learning with Explainable AI

Network traffic prediction is critical for ensuring the reliability and efficiency of communication systems. This study explores the integration of feature clustering and ensemble Machine Learning (ML) techniques to enhance the accuracy of network traffic prediction models. Using a comprehensive dataset created with the NS-3 Simulator, we evaluate multiple ML models, including Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), Decision Trees (DT), and Logistic Regression (LR). A Max Voting Ensemble (MVE) model incorporates the top three models and further refines the predictive performance. Comparative analyses are conducted with and without feature clustering, employing 10-fold cross-validation (CV) to ensure robust evaluation. Performance metrics, including confusion matrix (CM), are used to assess classification accuracy, precision, recall, and F1-score. The results demonstrate that the accuracy, precision, recall, and F1-score increased to 99.0, 99.14, 99.26, and 99.07, respectively, with the MVE model. Furthermore, we integrate Explainable Artificial Intelligence (XAI) techniques to provide insights into the importance of features and model decision-making. The results demonstrate that feature clustering significantly improves classification metrics, highlighting the potential of combining ensemble learning with feature engineering and XAI in advancing network traffic prediction.