An Improved Real UI-RBM Collaborative Filtering Recommendation Algorithm

The development of information technology has brought great changes to our work and life style.The popularity of Internet makes the information resources on the Internet show explosive growth.It is difficult to choose the valuable information for users,the utilization of information is very low.Therefore the screening and filtering of information has become a key issue,which is called information overload problem.At present,the most effective way to solve this problem is recommendation system,which has achieved great success especially in the field of e-commerce,Although the recommendation system has made great progress,there are some problems in the recommendation system.Data sparsity is one of the most important factors.In this thesis,we propose an improved real valued UI-RBM for collaborative filtering to resolve this problem.The main work is reflected in the following aspects:1.To solve the problem of data sparsity,we propose a matrix filling algorithm.The algorithm is based on the user collaborative filtering recommendation,the score prediction results are used to fill the sparse matrix.In order to explore the potential relevance of the user,to find a precise and effective neighbor,we proposed two new algorithms for computing user similarity in nearest neighbor queries: one is a conditional probability algorithm,the other one is a weighted value is added to the traditional cosine similarity formula.2.In order to deal with the problem of the real valued score and improve the accuracy of recommendation.The real valued UI-RBM model is proposed to recommend information to the target user.There are two differences between the model and the traditional RBM for collaborative filtering recommendation model: we model both user-user and item-item correlations in a unfied hybrid framework;the input unit is no a K two valued vector,but the real score.3.The dense matrix generated by filling the matrix is used as the input unit of the real valued UI-RBM model.We use Movielens dataset to test the performance of the algorithm proposed in this thesis and compared with the previous methods.The experimental results show that the proposed algorithm can improve the recommendation accuracy to some extent.