Bipartite network projection and its application in recommendation systems

How to get the weight between the nodes of the same kind of nodes without losing the information of bipartite network is the key question when projecting it. In this paper, the meaning correlation of nodes is studied by means of the information taking with the social tags, then, the gravitation between the nodes is calculated based on the Universal Law of Gravitation, the gravition equal to the weight between the nodes. After projecting bipartite network, users' preferences are spreaded on the correlation graph of resource using random walk method, its' steady-state probability on nodes of the graph can be gotten, the value reflect the degree of correlation between users' preferences and resource, the resource with higher steady-state probability are recommended to the user. the final experimental results proved this projection method can refecte the meaning correlation of the same kind of nodes, so the recommendation system can get better perfermanc.