User Behavior Analysis And Prediction Of Network Television

With the development of internet technology and the advance of triple play,Internet Protocol Television(IPTV)obtains the rapid popularity.For better comprehension of viewing behavior feature of IPTV users,this paper builds IPTV user behavior models from several angles.Via statistics analysis on 72 million real viewing records,we acquired many valuable results.Finally,a channel recommendation algorithm based on the user behavior analysis is proposed.Firstly we have shown the research on IPTV user behavior analysis and prediction at home and abroad,and have introduced the technology of IPTV system,user behavior analysis,and recommendation prediction.In the third chapter,this paper introduces and preprocesses the data set of IPTV user behavior,and puts forward the watching state transfer model orienting users—ISP.Then we classify the watching pattern according to the difference of watching duration and acquire IPTV user behavior cognition point—K1.Then we also classify the switching pattern according to the time interval of the two continuous two records and propose the rule of user state migration discrimination—K2.In the four chapter,we analyze the online users of the whole IPTV user group,hot channel,medium channel,and cold channel.We can allocate resources according to the variation characteristic of the online users,resulting in the maximization of the resource utilization rate.We discover the higherorder linear fitting function expresses the distribution of IPTV ratings more accurately than the Zipf distribution,providing highly available model for estimatting ratings more accurately.Through the analysis of the user flow characteristic: arrival rate,departure rate,online users and online users rate,we get a basic characteristic—K3,that IPTV users escape the channel together,and we can predict the switching moment according to this characteristic.Then through the compute and analysis of individual users,we find the consistency of the user watching behavior and the time,and the positive correlation of the user watching behavior and the user switching behavior.Through the characteristics of individual user behavior,we can describe the user portrait better,providing the basis for personalized precise recommendation.What's more,we has analyzed the channel switching delay from t hree angles: different users,different days,and different hours,which provides the basis for IP TV operators understanding QoE of users and perfecting IPTV system.Finally,according to the consistency of the user watching behavior and the time,in the five chapter,this paper proposes a recommendation method based on the user channel timeslot heat.In short,this research is the basis of the live streaming user behavior compute,and builds the new media user behavior basic theory framework based on the accurate ratings.This paper has the fundamental significance towards the pattern recognition of user behavior,intelligent recommendation,and preload,contributes to the improvement of resource allocation and IPTV QoE,and provides mathematics tool for the content and advertisement recommendation.