Research On Personalized Recommendation Of Family TV Programs Based On User Characteristics

Nowadays,with the rapid development of the Internet,the amount of data in the network is increasing exponentially.With the increase of data,digital television and communication technology have also developed rapidly,and the number of TV programs is increasing day by day.This leads to the difficulty for users to find the programs they are interested in in in in many TV programs.This situation will affect the viewing experience and television of users to a certain extent.The ratings of the program.Therefore,personalized recommendation of TV programs has become the common demand of more and more users and TV suppliers.Firstly,this paper analyses the characteristics and problems and challenges of personalized recommendation system for TV programs,and introduces the user preference model and the research status of personalized recommendation algorithm for TV programs at home and abroad.On this basis,the main recommendation system models and clustering models are analyzed,and the advantages and disadvantages of the derivative models of the two models are summarized.These studies lay a solid foundation for the improvement and design of personalized recommendation algorithm for TV programs.Then,this paper preprocesses the data in the user set-top box,cleans out the abnormal data in the data,and carries on the statistical analysis,which lays the foundation for further research.Then this paper constructs an implicit scoring system.On the basis of the previous scholars' implicit scoring based on user viewing time,this paper adds the important factor of user viewing rate,and constructs an implicit scoring system based on the combination of viewing time and video viewing rate.This system can further improve the interpretability of the algorithm and recommend the algorithm for the next step.Design and analysis provide a solid foundation for data security.Then,because the user scoring system is often sparse,the effect of personalized recommendation for sparse matrix will be discounted.Therefore,before personalized algorithm recommendation,this paper first recognizes the user's category features.Clustering algorithm is used to cluster users.Similar users are grouped into one group by clustering.When personalized recommendation is made later,personalized recommendation can only be made within differentcategories of users,which can improve the accuracy of the algorithm.When clustering users,in view of the inaccuracy of user distance measurement in kmeans +,this paper proposes a cosine similarity method instead of the original Euclidean distance,which can better measure user similarity.The results of an example show that the improved kmeans ++ algorithm has better performance in two test indices: intra-cluster distance and inter-cluster distance.Finally,this paper proposes a KNN collaborative filtering algorithm based on user rating feature optimization,which aims at personalized recommendation for users.Traditional KNN collaborative filtering algorithm only considers the common scoring features of users while neglecting the different scoring features of users.This method of evaluating similarity is unreasonable.Therefore,this paper proposes a user similarity calculation method based on the combination of user common features and different features,and then replaces the traditional similarity calculation method in KNN collaborative filtering algorithm.Finally,an example shows that KNN collaborative filtering algorithm based on user rating feature optimization has high recommendation accuracy,and performs well in MAE and RMSE.Based on the on-demand data of actual home TV users,this paper studies the personalized recommendation algorithm for home TV users,which has great practical significance,and can also provide reference for the research of personalized recommendation system for home TV programs.