IPTV User QoE Prediction based on Broad Learning System

Abstract

With the rapid development of Internet technology, users’ demands for Internet Protocol Television (IPTV) service quality continue to increase. Quality of experience (QoE) prediction has become an urgent problem to be solved because QoE is an important indicator for the evaluation of IPTV services. However, with the sharp increase in the amount of data, QoE is difficult to effectively predict using traditional machine learning methods. For deep neural networks, it may suffer from time-consuming training process due to a great number of parameters and complicated structures. In order to obtain the availability and efficiency of QoE prediction, this paper proposes an IPTV user QoE prediction scheme based on Broad Learning System (BLS). The designed BLS scheme contains three steps. Firstly, the collected IPTV dataset is preprocessed to remove abnormal data and useless fields. Secondly, network related parameters as well as user pause behavior, respectively belonging to objective and subjective influencing factors, are used for feature extraction. Subsequently, the extracted features are entered into the BLS for training, and finally the trained model is used to realize QoE prediction. Experimental results show that the proposed scheme is more effective and efficient for QoE prediction compared with competing models.