QoE-driven adaptive bitrate selection for cloud–edge 5G video streaming using Bi-LSTM and XGBoost

Abstract

The video over 5G networks is accelerating its rapid growth. Adaptive bit-rate selection mechanism intelligence is introduced to achieve the optimal user Quality of Experience (QoE) in dynamic conditions of the network. This paper presents a quick and intelligent QoE-based adaptive bit rate selection system for 5G video streaming with the help of cloud-edge that incorporates a hybrid approach of Machine Learning and Deep Learning. The study uses a Whale Optimization algorithm (WOA) to optimally fine-tune both Bi-LSTM and XGBoost hyperparameters to improve the model's performance, and our augmented hybrid system demonstrated much better performance in predicting the accuracy of QoE, accuracy in selecting the required bitrate, decreased bitrate switching, and less computational latency than the traditional ML models and isolated deep learning models. Experimental findings prove that the proposed method is always better than the baseline methods in terms of balanced accuracy, macro-F1 score, and QoE-relevant Top-2 accuracy, with more than 89% accuracy in Top-2 bitrate selection and low inference latency, which can be deployed on the edge. The results reveal that the developed cloud edge hybrid system can be used in real-time, QoE-aware adaptive video streaming in 5G-based networks. Our experiment has demonstrated that the system described in this paper can be deployed on the cloud-edge-based 5 G video streaming systems in real-time.