Personalized recommendation method based on dynamic multi-dimensional networks model

Personalized recommendations can supply resources to the interest preferences for users on the Internet. In this paper, we propose a personalized recommendation method applied in dynamic and multi-dimensional networks. By using this method, we are capable of predicting multi-directional relations of users on the Internet. Firstly, we present some algorithms to build multi-dimensional networks, reduction-dimensional networks and dynamic networks. Secondly, we cluster users by use of adjusted k-means algorithm. Thirdly, we get prediction ratings of objective user and do recommendation by dint of the nearest neighbors. Finally, we do experiments to test the correctness and efficiency of our method. The experiment results show that, compared with collaborative filtering recommendation systems, our recommendation system which utilizes algorithms of our work figures out less difference between prediction values and actual values, and the efficiency of recommendation system is improved to some extent.