QoE建模分析及会话流分类
Conversational flow classification based on QoE model and feature analysis

会话类流存在实时性高、时延抖动低、吞吐量大的特点,为了更好地对会话类混合流进行分类和提高用户的QoE（Quality of Experience）,文中从网络实时会话流入手，分析混合流的QoS(Quality of Service)并建立与QoE的模型关系，从而得出不同QoS特征下的QoE的概率分布，并把其中最明显的特征概率分布以网络流特征的形式加入到原有流分类特征集当中，然后对QQ视频、2D游戏、3D游戏、Skype语音四种会话混合流进行分类。通过C4.5决策树机器学习算法进行分类。实验结果表明：在流分类的正确率、召回率和F 测度上，使用QoE概率分布特征的分类算法相比于现有方案，性能均有显著提高。
Conversational flows have the characteristics of real time,low delay and big throughput.To better classify mixed flows and improve the user quality of experience (QoE),this paper analyzes real time flows based on the QoE model,and obtains the probability distribution function (PDF) of mean opinion score (MOS) value on different quality of service(QoS)parameters.Then,one of the most obvious characteristics of pdf are selected to add into the feature set,the mixed flows of the QQ video,2D game,3D game,and Skype are classfied.The paper uses C4.5 decision tree machine learning classifier.The results show that compared with the existing method,the accuracy,recall and F measure of the proposed method are significantly improve