Research On IPTV QoE Analysis Based On Data Mining

Mainstream telecom operators both at home and abroad are based on broadband IP network to carry out a large-scale IPTV business. However, with the number of users expanding and network congestion, IPTV service quality decline trend of increasing, increasing customer complaints, customer perceived quality(QoE, quality of experience) problem has become the major bottleneck restricting the development of IPTV business. On the one hand, the operator is difficult to take the initiative to obtain the quality of IPTV customer ratings, it is difficult to find the problem before the customer complaints. On the other hand, and IPTV quality related to network layer, business layer index number, operators also in daily operation collected IPTV related data, but it is difficult to determine QoE key indicators, is also difficult to from massive IPTV index data mining each IPTV customer daily perceived quality.Aiming at the above problems, this thesis includes three aspects:(1) from the existing IPTV related indicators, to find out the highest correlation with the QoE IPTV several key indicators;(2) the design of a QoE IPTV model and data mining framework, can get relatively accurate user QoE score from the mass of IPTV index data;(3) design a set of Qo E IPTV evaluation prototype system based on data mining. This thesis works as follows:First of all, this thesis investigated the IPTV related indicators in the industry, the existing QoE IPTV evaluation methods, data mining algorithms, and QoE evaluation method based on data mining research and analysis.Secondly, the IPTV parameters are divided into three categories: IPTV video QoE parameters, IPTV bearer network QoE parameters and user terminal performance QoE parameters; Using factor analysis method, principal component analysis and correlation analysis method to find out the key indicators affecting QoE and QoE key indicators of the mining method and the existing index acquisition methods are analyzed and compared. Finally, we give the IPTV QoE index system.Thirdly, a QoE evaluation method is proposed, which is based on the fitting method, which is used to construct the calculation process from the original index to the QoE score. The evaluation algorithm is realized from the original IPTV index data to the QoE IPTV score of the mining output.Finally, this method is used to design a set of QoE IPTV evaluation prototype system, in which the simulation experiment and real data set are two ways to verify the validity of the prototype system and the evaluation method. At the same time, using the goodness of fit of the regression analysis to verify the QoE fitting function in this thesis can reflect the relationship between the index and the QoE.