Research On The TV Program Recommendation System Based On The Family User Model

In recent years,radio and television technology changes with each passing day,and TV programs are increasingly abundant.While the number of TV programs increases,it also makes it difficult for users to efficiently choose the TV programs they are interested in,which also affects the audience rating of TV programs to some extent.Traditional TV services have been difficult to meet the demand of family users.At present,it has become a common demand of cable network operators and TV users to recommend TV programs they are interested in.And it is also one of the important issues in the field of recommendation.Today's mainstream user behavior analysis models are mainly applied to E-commerce websites,mobile news,OTT and other platforms.Traditional Internet resource recommendation system is mostly recommended for individuals,there are relatively few models of multi-person complexes with different ages,different time periods,and different program preferences for family users.As for the recommendation algorithm,they are mainly divided into several categories,such as popularity-based,content-based,collaborative filtering,mixed recommendation and so on.They each have their own style.And they also their own advantages and disadvantages.Therefore,this topic is based on the family user model to help a broadcasting and television network operator conduct research and development of the cable TV recommendation system.Through the review of related literature at home and abroad,the study of TV program recommendation system characteristics,difficulties and challenges are analyzed,and further understand the recommendation system modeling in the user's interests and family oriented user recommendation system algorithm,the current status of and the corresponding typical recommendation algorithm and the introduction of evaluation index,in order to further optimize modeling for user's interests and home users recommendation algorithm research,laid the theoretical and practical basis,thus determining the overall design of cable TV recommendation algorithm.Aiming at the characteristics that cable TV program personalized recommendation is usually targeted at family users with diverse interests.This paper proposes a time-sharing user interest preference modeling method based on viewing behavior and program characteristics.The method firstly analyzes and models the user report,calculates the similarity of the user in the time domain and the similarity of the program,and clusters the user and the program,and finally obtains the user and the program classification model in the time domain.Then,according to the time-sharing model,based on the different characteristics of the TV program and their impact on time,the viewing behavior of users is separately quantified.Thereby,a family user model for television program recommendation is obtained.On the basis of the time-sharing user model and family members' speculation,the family members' different viewing habits for various programs are further obtained.Aiming at the large data of user ratings and the sparse interest preference matrix of the broadcasting and television network operator,this paper proposes a recommendation method combining offline recommendation and online real-time update.In the offline recommendation process,the collaborative filtering algorithm based on matrix factor decomposition is adopted for TV programs with viewing behavior records.The algorithm is based on the implicit semantic model,which uses the least squares method to realize matrix decomposition,and limits the parameters such as the number of iterations and implicit semantics.At the same time,the regularization term with ? value is added to prevent the over-fitting of the proposed algorithm model.The collaborative filtering algorithm based on content is adopted for TV programs without viewing behavior records.For advertising resources,knowledge-based recommendation is adopted.Finally,the current users are identified based on the real-time viewing time of current family members.And,the current TV programs which current family members are interested in,as well as advertisements are pushed to the current.Using the original audience behavior data of a broadcast network operator,the test of the cable TV program family user model and recommendation algorithm proposed in this paper,including recommendation accuracy test,recommendation accuracy convergence test,processing time test,and recommended coverage test.Theexperimental results show that the proposed method can effectively identify the interest preferences of cable TV users,and improve the performance of TV broadcast system of a broadcasting network operator compared with the recommendation accuracy.