A Proposed QoE Prediction in Video Streaming Using ITU-T Standards and Machine Learning Approaches

The Quality of Experience (QoE) has become a crucial research topic for network operators and video providers, since it directly measures customer satisfaction.Objective QoE Assessment outperforms subjective in cost and applicability.However, predicting the QoE remains a challenge due to the variety of its influencing factors.The P.1203 by the ITU-T has emerged as the first QoE standard in video streaming environments.This work aims to predict the next possible QoE degradation to avoid poor perceived quality.A low user interaction is simulated using Selenium to extract the ITU-T P.1203 video parameters.The obtained data, in addition to network data from a network prob, have been used to train Feed Forward Neural Network (FFNN) classifiers and several regression models.The QoE has been predicted using both classification and regression in three implementations, in addition to implementing feature selection for better feature space size and the prediction performance.The FFNN results have been evaluated using cross-validation accuracy, RMSE, and confusion matrices.Combining the ITU-T P.1203 standards with the advances of machine learning (ML) in this study provided high prediction accuracy levels that exceeded 88% and highlighted nine selected influencing factors that are highly impacting the delivery of video services.