The recommendation systems for smart TV

In this work, our recommendation system aims at developing a new user friendly technology for smart TV based on control TV by voice command and intelligent recommendation. For this recommendation system, we solve three issues: the realtime data of TV programs lists collection, the design of the low response times and high concurrency server, and the recommendation TV programs which users interested in according to their preferences. The real-time data of the TV programs fetches from the internet by the web crawlers which are working with a concurrency approach, but the data is not in a standard format or encoding, therefore it is necessary to convert these variety of data into well-structured format and store in the database before user selection. In addition, the thread pool technology is used for the high concurrency. At last, with the purpose of recommending the individual TV program for a user, a novel hybrid collaborative filtering approach based on statistic and affinity propagation is proposed, which can automatically analyse user's favorite TV programs and recommend the ones with more interest. The experimental results have shown the proposed method obtained good performance which the average MAE for uses is 0.1585.