Research On Collaborative Filtering Recommendation Algorithm Based On Implicit Feedback For IPTV Users

With the promoting of three net fusions,Internet Protocol Television(IPTV)has been used by millions of families.Its massive video resources and high quality of services attract a great many of customers.However,the rapid growth of the information-carrying capacity of the network lead to the explosion of the amount of video resources on IPTV.This will not only cause some difficulties for users to find their favorite programs,but also have a bad impact on the user's satisfaction of IPTV.At this time,IPTV recommendation system will serve as a very valuable tool to improve this problem.However,compared with the conventional TV,IPTV provides multiple types of service and a large number of programs.When users select a program,they need to switch between interfaces constantly.In order to reduce as much as possible the amount of user's operation,selecting appropriate scenes to present recommendation results is necessary in IPTV systems.Also,in order to provide users with personalized recommendations,recommender systems need to collect information about user's preferences.Since users are reluctant to invest much time in explicitly expressing their interest,preferences often need to be implicitly inferred through data gathered by monitoring user behavior.Generally,it requires a user interest model to describe these user preferences as a whole.The user model directly affects the quality of the recommendation result,and a good user model not only need to combine the characteristics of both systems and user data,but also need to satisfy the requirement of the recommendation algorithm used in the system.To solve above problems,this paper propose collaborative filtering algorithms based on implicit feedback for IPTV users.The main results of this study are: 1,learn an implicit user model by tracking the programs users have watched.In this process,this paper firstly analyzes the features of IPTV users and videos to be recommended in the system.Then it propose to establish an user-item matrix as the user interest model for IPTV recommendation system.Finally,this article develop a detailed and reasonable rates implicit policy for building the user model.2,provide users with recommendation result on different IPTV terminal interfaces.This paper firstly select two scenarios in different IPTV terminal interfaces.Then according to the needs of different scenarios,it choices reasonable collaborative filtering recommendation algorithms and designs recommendation programs for them to achieve recommendation results to users.3.achieve the video recommendation results for the two different scenes and analysis the experimental results.