Personalized Recommendation in Dynamic and Multidimensional Social Network

Abstract

A personalized recommendation algorithm applying in dynamic and multidimensional network was proposed in this paper. First, built up multidimensional network for users on the basis of user's model and rating collection, and formed dynamic and multidimensional network by combining with local-world evolving theory. Then the algorithm clustered users by making use of adjusted k-means algorithm. Finally, got the prediction ratings of objective user and did recommendation by dint of the similarity between neighbors. Comparing collaborative filtering and content-base recommendation system, the experiment result shows that the recommendation system which utilizes this algorithm, figures out lower MAE and higher recall and precision separately. That is to say, the quality of personal recommendation has been improved to some extent.