Research Of Collaborative Filtering Recommendation Algorithm

With the rapid development of information technology, the information on the Internet shows explosive growth trend. And the massive data result in the information overload problem. How to help users find the useful information from the mass of data attracts attention of academia and industry. As an effective mean of information filtering, the recommender system has been widely used in various fields. Recommendation algorithms as the core of recommender system are gaining more and more attention from researchers. Collaborative filtering (CF) is one of the most popular algorithms. However, it also has some shortcomings, for example, low recommendation accuracy, cold start and poor scalability.In this paper, to improve the accuracy of recommendation algorithm, we carried out some researches on collaborative filtering algorithm in the following aspects of:Firstly, the traditional matrix factorization (MF) algorithm generates recommendation using rating information but ignoring the content information. So, we proposed an improved MF model. This model considers not only the rating information but also the content information. Secondly, traditional neighbors-based collaborative filtering algorithms always predict ratings based on the weighted average method, but this method will result in the prediction bias. So, we carried out some researches on the combination of matrix factorization and neighbors-based algorithm, proposed the adjusted neighbors-based algorithm. Thirdly, an algorithm combining matrix factorization, users-based collaborative filtering and items-based collaborative filtering together was proposed to improve the prediction accuracy. This combination algorithm takes advantage of global characteristic of the matrix factorization algorithm and locality characteristic of neighbors based algorithm.Finally, we performed experiments on MovieLens dataset and Baidu movie datasets. And the experiment results show that the proposed methods improve the accuracy of the recommendation algorithm.