An urban traffic signal control network partitioning method using Self-organizing Maps (SOMs) is proposed in this paper to divide the urban road networks for signal control coordination. Two compound parameters, saturation degree and travel time are selected as the model inputs to divide the network into several sub-networks for the purpose of traffic signal coordination. According to the network division principles, the proposed model visually explores the traffic pattern of the road network and shows the SOM’s neural weight distribution can represent traffic patterns of a network and could be well trained to cluster intersections with similar traffic patterns. Comparisons of the network partitioning performances for Chuzhou city of different initial SOMs network topologies and hierarchical clustering model are also discussed. The results of the experiments indicated that the proposed model could capture the traffic pattern of the road network and reflect a real network division situation. The proposed model can be used to directly divide urban road network based on real-time traffic information for traffic signal coordination.