Study And Application On An Improved Recommendation Algorithm

Now, we are in an era of “information overload”. How to accurately find information and find people's need has become a problem when facing of the vast amounts of information on the Internet. The recommendation system is a powerful tool to solve this problem. Recommendation system predict user's preferences, and recommend to the user what he is interested in by analysing of user's behavior records, constructing the user's interest model.Collaborative filtering is the most popular method in the recommendation system at present. It find the connections between the users and the items by analysing of the historical behavior records of all users in the system. Collaborative filtering is divided into algorithm based on neighborhood and the latent factor model(matrix decomposition model).After the Netflix Prize, Matrix decomposition model become a hot research topic in the recommendation algorithm, all sorts of the improved model are emerge in endlessly. In this article, we introduce the traditional matrix decomposition model, introducing the implicit feedback model, and matrix decomposition model based on the time effect. Time effect is a deep research on the current recommendation system, unlike previous static recommendation system, by adding time effect, rating recommendation system can predict more accurately.In the matrix decomposition model based on the time effect, we put forward by the time function to optimize the training parameters of time. By this means, reduce the time and space complexity of the algorithm. At the same time, we combine data normalization and matrix decomposition, further optimize the accuracy of the algorithm.Finally, we designed and implemented a campus video sharing platform based on the improved recommendation algorithm. By recording the various operating behavior of users, we set up the user's interest model and provide personalized recommendation service to user.This helps user to find their interested video more effectively and quickly, improve the user's experience.