A recommendation system and method for measuring node intimacy in a weighted graph with direction comprises the followings: (1), obtaining the weight of an edge in a given graph: giving a definition of weight for the edge in the graph according to customer requirements and weighting the weight of the edge according to definition; (2), carrying out normalization treatment on the given weight of the edge in the graph: after the weight of the edge is obtained, carrying out normalization treatment on the weight of the edge, so as to calculate the intimacy between directly-connected nodes; (3), obtaining importance measurement of the nodes in the given graph: calculating the importance of nodes in the graph, under the same topological structure, the intimacy of an important node and a given node can be higher; (4), adding a new virtual node into the given graph: adding an edge between the virtual node and the given node, and appointing weight to the edge according to node importance; (5), calculating the intimacy results among final nodes: calculating the intimacy between the appointed node and other nodes by using a restart type random walk algorithm on a weighted graph alternated.