Software Defined Network Traffic Classification for QoS Optimization Using Machine Learning

Abstract In the era of rapidly expanding network infrastructures, ensuring optimal performance and quality of service (QoS) for diverse applications face significant challenges. Traditional traffic classification (TC) methods often fall short due to their inability to adapt to the dynamic and complex nature of modern network environments. To address this limitation, this paper proposes integrating software defined network (SDN) architecture with machine learning (ML) technology. The study examined four scenarios: multiclass classification and binary classification, both before and after scaling. We used various ML models, including linear, non-linear, and hybrid models. To evaluate the performance of these models, we utilized several evaluation metrics, such as accuracy, F1 score, kappa score, ROC curve, and confusion matrix. The paper examined different feature scaling methods, including standard scaling, min-max scaling, max-abs scaling, and robust scaling. The results showed that both min-max and max-abs scaling provided the best performance enhancement across the four scaling methods. Finally, XGBoost model provided the highest performance across all scenarios, with accuracy reaching up to 99.97%.