Design And Implementation Of Recommender System Based On Collaborative Filtering Algorithm

The current rapid development of Internet technology provides a majority of users with a lot of data which meet the user demand for information. However, a variety of overwhelming information lead to a fact that more and more users feel difficult to search their interesting information, with the phenomenon of so-called "information overload". The personalized recommendation technology is an effective method to solve this problem. It recommends possible information which users might be interested in, by analyzing historical information about the user to establish the interesting model. It needs no any inputs from the users. The collaborative filtering algorithm is not only the most mature, but also the most widely used technique in the research field of the recommendation technology. In this paper, we made some elaborate analysis and discussions about the collaborative filtering algorithm, designed a hybrid collaborative filtering algorithm, and built a recommendation system of movies which based on that algorithm.The main works of the paper as follows:At first, analyzing three collaborative filtering algorithms which based on users, projects, and Slope One separately. The collaborative filtering algorithm based on Slope One has some advantages that it is has high efficiency, easy to implement, and higher accuracy of the recommendation results. But it does not handle the personalized recommendations between different users well. After that, we mentioned a collaborative filtering algorithm which combined user similarity with Slope One algorithm. The core idea of the algorithm is to calculate the average similarity between users, and predicted scores by using similarity among users as weights of Slope One algorithm. Finally, recommending to the users by the resulting scores which came from the prediction. Taking into account the user’s neighborhood search strategy, the paper also designed a dynamic threshold to find the user’s nearest neighbor method. To solve the problem of computational complexity when huge amounts of data, we use the distributed computing to design the hybrid collaborative filtering algorithm. We could see that the collaborative filtering recommendation algorithm will improve the accuracy of personalized recommendation system after theoretical analysis and several experiments based on the standard MovieLens dataset.In the end, after investigated from the demands of the movie recommendation system, and designed the architecture, module, and database based on the previous foundation of research, we finally implemented the hybrid collaborative filtering recommendation system.