An Adaptive Hybrid Ensemble Learning Framework for Quality of Experience Prediction in 5G Video Streaming

Content streaming applications in fifth generation (5G) wireless networks are growing more popular, leading to stronger demands for more precise Quality of Experience (QoE) prediction to guarantee quality delivery of multimedia services. Traditional machine learning algorithms typically rely on a plethora of Quality of Service (QoS) parameters, which increases the complexity and inefficiency due to the redundant parameters in prediction tasks. In this paper, we put forward an Adaptive Hybrid Ensemble Learning Framework (AHELF) by integrating data preprocessing, adaptive feature selection, and ensemble classifier to accurately predict the QoE in 5G video streaming scenarios. The collected network and video specific parameters are first pre-processed. Afterwards, two feature selection techniques, i.e., Correlation analysis and Recursive Feature Elimination (RFE) are applied to identify the significant features that contribute the most to QoE prediction. The selected features are then used in a hybrid ensemble learning model, composed with three different classifiers Random Forest, Gradient Boosting and Extreme Gradient Boosting, which based on the Weighted Decision Fusion Technique. The suggested framework is evaluated on a 5G video streaming dataset, and the experimental results shows this framework outperforms traditional machine learning algorithms in terms of accuracy, precision, recall, F1-score by obtaining 98.31% of accuracy.