An improved clustering-based collaborative filtering recommendation algorithm

In the circumstance of big data, the traditional collaborative filtering recommendation algorithm in e-commerce system is faced with the problems of data sparse, accuracy, real-time and etc., this paper proposes an improved clustering-based collaborative filtering recommendation algorithm. The algorithm introduces time decay function for preprocessing the user's rating and uses project attribute vectors to characterize projects, user interest vectors to users and use clustering algorithms to cluster the users and the projects respectively. Then the improved similarity measure methods are used to find the user's nearest neighbor and project recommended candidate set in the cluster. Finally, recommendations are produced. Theoretical analysis and experimental results show that the algorithm not only can effectively solve the problems of data sparse and new project, but also can portrait for users in multi dimension and reflect the user's interest changing. The recommended accuracy of the algorithm is improved obviously, too.