Data-driven QoE prediction for IPTV service

With the development of Internet and multimedia technology, more and more families enjoy smart multimedia services provided by Internet Protocol TV (IPTV). It is crucial for operators and content service providers to find key indicators and improve quality of experience(QoE) for users. In this paper, we propose data-driven QoE prediction for IPTV service. Specifically, we define QoE to evaluate user experience of IPTV in data-driven approach at first. Then we analyze user's interests and device indicators to understand when and how they affect user experience. Based on user interest lists for both regular users and new users, we propose the uindex to quantify user's interests in Live TV. Finally, we build a personal QoE model based on an artificial neural network (ANN). Experimental results show that uindex improves the integrity of QoE description. Moreover, the model can predict QoE with an accuracy of 83.93% for regular users and 83.90% for new users in the record level, better than those of competing algorithms.