Design And Implementation Of IPTV User Experience Prediction System Based On Neural Network

In recent years,with the rapid development of Internet technology,interactive network TV(IPTV)services have become increasingly diversified,which is main entertainment way for users at home.In order to improve user retention,IPTV operators and service providers are paying more and more attention to user quality of experience(QoE).The perception and evaluation of QoE have become a hot research issue for industry and academia.The traditional way to get QoE by user scoring is timeconsuming and labor-intensive,and the real-time performance is not enough.It is not suitable for the evaluation of users QoE by IPTV operators and service providers.Therefore,it is urgent to establish a QoE prediction system,which can improve the service and ultimately improve user QoE.Based on this,this paper designs and implements a neural network-based IPTV user QoE prediction system in a big data environment.The main research works and innovations of this thesis are listed as follows:Firstly,this paper propose a hybrid feature extraction method based on subjective and objective factors.On the one hand,from the perspective of correlation analysis method and information gain,the relationship between objective indicators and user QoE is analyzed,and the influence of redundancy factors on subsequent modeling is removed,and the objective characteristics affecting user QoE are selected.On the other hand,from the perspective of user behavior,the characteristics related to user behavior are analyzed in depth,and subjective indicators affecting user QoE are extracted.Compared with other feature extraction methods,the hybrid feature extraction method proposed by the thesis is more comprehensive,which can lay a solid foundation for subsequent prediction.Secondly,this paper design a QoE prediction algorithm for IPTV users based on neural networks.Considering the long-term and short-term dependence of user QoE,paper chose the long short-term memory neural network(LSTM)to model user QoE.Further,for the limitations of the LSTM algorithm,the LSTM-Attention model is established in this paper.By introducing the attention mechanism,the model accurately describes the contextual association of the user QoE and improves the performance of the original model.The experimental results show that the prediction accuracy of the model is 7.8% higher than the existing mainstream algorithms.Finally,based on theoretical research,this paper designs and implements an IPTV user QoE prediction system.Specifically,based on big data core technology and visualization technology,Echarts is used as visualization framework,Spark is used as distributed parallel computing framework,supplemented by Streaming as real-time stream computing.In addition,SparkML is selected as the machine learning library,and SparkSQL is used for interactive analysis.With YARN as the resource manager,Zookeeper performs distributed service coordination to ensure the system is stable and efficient.The experimental results show that the IPTV user QoE system can realize real-time accurate analysis and prediction of user QoE.