Design And Implementation Of Video Recommendation System Of Smart TV

In recent years, with the rapid development of Internet technology and related applications, especially the rapid development of social networks, a large amount of information is generated all the time, which makes users have to deal with the problem of information overload. Users find it difficult to seek their own interest or useful information when facing the massive data on the internet. With the diversification of the generation of network video, video websites are also facing the problem of information overload, which makes content providers difficult to push the high quality content to users accurately. Recommendation system is one of the effective methods to solve these problems. It analyses the history of user’s behavior and social information, and then establishes user’s interest model, which is used to predict the future preferences of users, and to recommend the information that he might be interested in or thought useful.Traditional video recommendation systems are often used in the film and video web sites. It can help users to find out films or videos they are interested in with massive video contents. With the invention and rapid development of Smart TV, users can get massive network videos by using related software on smart TV at the same time they are watching TV programs. However, we should both recommend network video and live TV shows, so it is a new research direction how to relate the traditional radio and television network and realize the personalized recommendation of the traditional broadcast TV networks and the Internet multimedia data.This dissertation uses the Spark framework, which is a general distributed parallel computing framework developed by UC Berkeley AMP lab. Besides we combine the Nosql-Mongdb and the distributed storage system-HDFS to design and implement the video recommendation system of smart TV. The system establishes user’s interest model by analyzing and processing a large number of viewing behavior data of users collected by smart TV and then recommend videos that they like.Recommendation algorithm is the most important part of the entire recommendation system and it determines the merit and demerit of the system significantly. This dissertation proposes a collaborative filtering recommendation algorithm based on the time period. Because different groups watch TV at different times, we divided one day into several periods and then accomplish a video recommendation algorithm adapted to smart TV based on text classification algorithm k-means and objects recommendation algorithm.This dissertation describes the design of the smart TV recommendation system framework, and then describes the design and implementation process of the algorithm in detail. Finally, the performance and accuracy of the system are tested. Test results show that the recommendation system can work properly and it can adapt to the smart TV terminals.