Design And Implementation Of The Program Recommendation System Combined With Creators Information

With the popularization of IPTV and smart TV and the rapid growth of TV program content,people have moved from the era of general lack of content to the era of excess content.Faced with a large number of programs,it’s hard for users to find the content they are interested in,leading to the problem of information overload.Therefore,how to actively recommend programs of interest to users is a very important research topic for both viewers and operators.The large-scale use of two-way network TV set-top boxes allows a large number of users’ viewing data to be collected,which provides an important foundation for mining users’ preferences with big data technology.On the other hand,program creators have gained more exposure in the current era of mobile Internet,and the dissemination of TV program content has entered the mode of fan idols.The factor of program creators is becoming more and more important to the audience,much more important than other factors such as program type.In this context,this thesis designs and implements a recommendation system based on user viewing data and combining the main creator factors.The main research contents of this thesis are as follows:1.A program recommendation method based on the main creator information is proposed.First of all,based on the viewing data,this thesis models user preferences from the viewing dimension,time dimension,and creator dimension.The creator dimension considers the different influences of the main creator’s division of labor,the main creator’s popularity,and the main creator’s CP.Then,the collaborative filtering algorithm is used for recommendation.Considering that in the scenario of TV program recommendation,the collaborative filtering algorithm has problems such as sparse scoring matrix,low accuracy of recommendation results,and cold start of programs.This thesis proposes a method of filling the rating matrix with the similarity of the main creator to solve the sparsity problem of the rating matrix,and integrates the similarity of the main creator in the process of calculating the similarity of the collaborative filtering algorithm to solve the problem of cold start of the program.Finally,the algorithm performance test is carried out.This thesis uses the original viewing data of a certain operator as the experimental data set,compares various experimental schemes,and uses the root mean square error(RMSE)as the score prediction accuracy index.The experimental results show that this thesis is in the recommendation process.Compared with the traditional collaborative filtering algorithm,the introduction of the main creation information effectively improves the accuracy of the recommendation.2.Completed the design and implementation of the system.First of all,starting from the actual application requirements of the system,clarify the functional requirements and non-functional requirements of the system.Then,the overall structure of the system is designed based on the actual demand,and the functional modules of the system are divided.Finally,each functional module is implemented one by one.3.Complete the system test work.First the system test is carried out.In this thesis,the functional test and non-functional test of the system are carried out by designing reasonable test cases.The test results show that the various functions of the system can run normally,provide performance guarantee and meet the non-functional requirements of expansibility and data security.