Machine Learning-Driven QoE Prediction for SDN Networks: A Hybrid Model Approach

Predictive accuracy of Quality of Experience (QoE) is crucial for maximum user satisfaction since cloud computing and video streaming and real-time communication heavily depend on high-speed networks. The Quality of Experience-aware traffic control method known as SDN enables programmable network management strategies. Majority of modern research about QoE prediction focuses solely on packet loss and delay measurements without providing an expanded view of network operation quality. The proposed study implements QoE prediction by adding latency, packet loss, jitter, and throughput data from the Mininet software and ONOS controller functioning under SDN configurations. The collected data served as an input for training Random Forest, XGBoost and Gradient Boosting models through assessments of Pearson Correlation and R2 score with the additional metrics Mean Absolute Error and Root Mean Square Error along with cross-validation accuracy. The introduction ofa set of combined algorithms boosted the prediction accuracy significantly. The combination of separate models achieved superior performance in the experiment by reaching a validation accuracy rate of 95.36% Quality of Experience (QoE) is a customer-centric metric that captures the extent to which the network is performing from the point of view of customer satisfaction, as compared to Quality of Service (QoS), which is related to network side metrics like latency, jitter, packet loss, and throughput. While traditional QoS metrics are informative about network conditions, they tend not to map well to the end-user's quality of experience, and hence accurate prediction of QoE is an important part of network optimization.