Accurate and quick decision of emergency medical services (EMS) based on the locations of patients and hospitals is important for life saving. This research regards the correspondence of the locations of patients and hospitals as a bipartite graph structure, and proposes a method for forcasting frequencies of the conveyance from a certain location of patients to a certain hospital. Experimental results using real EMS data of Tokyo metropolitan show the effectiveness of our method. This is a joint research of Tokyo Institute of Technology, National Institute of Advanced Industrial Science and Technology and The University of Tokyo. And this is a part of the outcomes of Research topic 2-5 "Learing and mining graph data and their applications" in AIST-Tokyo Tech Real World Big-Data Computation Open Innovation Laboratory.