Abstract: To explore the potential traffic characteristics of large-scale and complex networks, it is difficult to reflect the spatial characteristics of traffic flow only by considering that the total traffic volume is not reflected, and the intensity distribution of traffic flow in different OD directions should also be considered. So, the paper proposes a network partitioning method that considers the OD directions of traffic subregions, thereby better reflecting the intensity distribution of traffic flows in different OD directions within the traffic subregions. Firstly, with the dual objectives of minimizing the variance of vehicle speeds within the initial region and initial region compactness, the original network is partitioned into relatively large initial regions. On this basis, with the objectives of minimizing the vehicle speed variance within regions, enhancing network compactness, and balance the scale of links in each region, a mathematical optimization model is established to merge the initial regions. Technique for Order Preference by Similarity to an Ideal Solution is used to assess these three objectives, and a better partitioning scheme is obtained by solving the model with a genetic algorithm. Then, vehicle GPS data is processed to obtain vehicle trajectories. Based on these trajectories, the speeds and link IDs of the link of region along each vehicle's OD path are prestored. A method for calculating the speed variance of regions based on vehicle trajectory data is proposed. Finally, the GPS data of buses in Shenzhen is used as a case study to verify the effectiveness of the proposed partitioning method. The results show that, compared with traditional methods, the proposed method achieves more compact and uniform regions, and can improve the partitioning efficiency. Moreover, by constructing a macroscopic fundamental diagram(MFD) that the OD directions of traffic subregions, the spatial evolution characteristics of traffic flow in the network can be better captured.