Rating prediction algorithm and recommendation based on user beahavior in IPTV

Service quality of IPTV directly influence Quality of user´s Experience (QoE), one of the key technologies to attract new users. The current researches of IPTV mainly focus on two aspects: On one hand, researchers are concerned on evaluation of the quality of videos; on the other hand, personalized recommendation is cared more and more. For the former, the most effective solution is to improve the bandwidth of IPTV network; but to the second, Collaborative Filtering (CF) Algorithm performs perfect effect in personalized service. This paper we mainly pay attention to the later, based on the interests of user. Owing to the characteristic of interactions between user and television in IPTV platform, different behaviors of user, such as explicitly rating behavior, watching behavior and saving behavior and so on, may show different interests of Items. To obtain interests of user and make Personal recommendation, the author firstly introduced related behavior mining algorithm according to the main three behaviors and then proposed a new similarity computation in recommendation based on CF. Finally algorithm performance is evaluated with modified IPTV data from real TV watching data provided by Wenguang Shanghai Corp. in China and it shows quite comparative quality of recommendations.