Data-driven objective evaluation on IPTV user experience

The emergence of network television (IPTV) has heavily promoted development of the technology and prosperity of the market in the TV industry. In order to win in the fierce competition, IPTV content providers need to be sure to provide users with interesting service, and operators also need to ensure quality of service transmission so that they can improve users quality of experience (QoE). How to use the objective method to evaluate user subjective characteristics is critical. In this article, we study the IPTV user experience by data-driven objective evaluation algorithm to solve the problem. First, the set-top box data collected from ZTE are pretreated; then we establish the model which map from network situation to viewing time. We perform a new regression algorithm which combine weighed-KNN with classification and regression tree (CART). Compared with the traditional KNN, experimental results show this algorithm can improve the correlation coefficient, reduce the prediction error, and alleviate testing time.