Research And Implementation Of TV Program Recommendation Based On IPTV

In recent years,smart TVs and digital TVs have become one of the most important consumer products for the Internet of Things.The number of television shows has greatly increased.Compared with traditional TV programs,the advantages of smart TVs are reflected not only in real-time watching the same real-time online programs with traditional TVs,but also can through the data platform provided by service providers to search for the programs you want to watch online.However,the emergence of a large number of TV programs has also made it difficult for users to accurately and quickly search for their favorite TV programs in massive TV programs.As a result,some wonderful TV programs have not received the proper ratings,which has led to suppliers cannot accurately provide corresponding video programs and resources.Therefore,intelligently recommend TV programs of interest to users has become a common and urgent need for TV service providers and users,and it is also one of the important problems to be solved in the field of recommendation.First of all,it introduces the characteristics of TV recommendation system and some challenges it faces,the current research status of recommendation system and TV program related recommendation system,the mainstream recommendation technology stack and its related algorithms.Then it analyzes the relevant information of user log provided by the existing city TV station,stores and cleans the relevant data.Next,we get the metadata of the program information through the web crawler on the Douban.com,and introduce the related methods of the type classification of the program.Then,an improved user scoring model is proposed based on the time per viewer of the IPTV user.The main purpose of the scoring model is to solve the problem that the IPTV user rating is difficult to obtain,thereby improving the accuracy of the recommendation.At the same time,for another major problem of IPTV recommendation,the division of family members,a method based on K-means to dynamically divide family members is proposed.The family member is dynamically divided by the time distribution of family members watching TV.The purpose is to be more targeted when recommending programs,and reduce the chance of recommendation errors.Finally,the design experiment compares the results of different models,makes use of today's more mature technology stack,adopts a scalable distributed technology solution,and verifies the effectiveness and feasibility of the recommended method by comparing the speed and accuracy of the method.Through the comparison of different models,it is proved that the method of dynamic division of family members based on K-means proposed in this paper is very suitable for IPTV recommendation,and it has good expansibility and ease of use,and is very suitable for the use of production environment.