Research Of Video Recommendation System Based-on Hadoop

With the development of the Internet era, people's entertainment life has been greatly enriched in recent years. Every day,there are a lot of video on the Internet. This is an urgent problem to find an answer, that people find the information to meet their demand.For this reason, there are a lot of research and solutions, which recommended the system is an excellentsolution.Recommendation system is an effective technical scheme.Recommendation system collect user information and user history information to predict the user preferences for video, which filters out a large amount of video information.Currently, the main factors affecting the development of the recommended system is a data sparse, cold start and magnitude expanding existing resources, etc.We can improve the accuracy of collaborative filtering algorithms, in order to solve the recommendation system in the development of these problems. In the vast amounts of data,we use Hadoop to build reliable, high-availability of large data processing platform.There are a detailed description of the structure and working principle of the two components of Hadoop in the paper. Finished recommendation system in large data processing platform, with good large data processing capability,it provides follow-up studies with a new research direction. The following contents are the main content of the paper:1)Recommendation algorithm analysis which are used by recommendation system are researched.And I summarize their advantages and disadvantages.And we proposed the concept of a mixed recommendation,which is based on association rules recommended and collaborative filtering recommend.2)The solution of user cold start:The recommended method use registration information to provide personalized recommendation.The recommended method provides hot items for anonymous user.3)The solution of item cold start:we can use content information items, establish weighted vector model, and seek similarity items.4)Improvements to the data sparsity problem:In practice,we can distinguish between different collaborative filtering algorithms,selection of appropriate algorithms to solve the problem of data sparsity.5)The video recommendation system can build on Hadoop platform, and we describe the frame design, module design, recommendation algorithm design and database design in the recommendation system.Finally, combined with the specific experiment,we compare hybrid recommendation algorithm with collaborative filtering algorithm based on item.According to the evaluation criteria recommended by the algorithm,the algorithms are respectively compared in the data of MovieLens.