Using Advanced Ensemble Machine Learning Models to Predict Traffic in SDN-Based Networks: A Comparative Study of Bagging, Boosting, and Stacking Approaches

Predicting network traffic in Software-Defined Networking (SDN) environments is essential for proactive resource allocation and congestion management. Ensemble machine learning models that combine multiple weak learners, including Bagging, Boosting, and Stacking, have demonstrated superior predictive capabilities in complex domains. This study evaluates the effectiveness of Random Forest (RF) as a Bagging method, XGBoost and LightGBM as Boosting techniques, and a meta-learner-based Stacking ensemble for traffic prediction in SDN networks. Using a comprehensive dataset of SDN traffic traces, we investigate model accuracy, training efficiency, and robustness. Experimental results reveal that Stacking leveraging base learners from both Bagging and Boosting families consistently outperforms individual methods, offering a balanced trade-off between accuracy and computational cost. This work highlights the potential of ensemble strategies for dynamic traffic prediction and intelligent SDN management.

کلیدواژه ها:

SDN ، Traffic Prediction ، Ensemble Learning ، Random Forest ، XGBoost ، LightGBM ، Bagging ، Boosting ، Stacking