Large seagoing vessels experience substantial changes in navigational conditions when transitioning from open sea navigation to port connected and restricted waters. Owing to their large dimensions and maneuvering characteristics, their operations are particularly sensitive to local navigational constraints, vessel interactions, and surrounding traffic, resulting in spatially heterogeneous traffic responses. This study proposes a data-driven approach for spatially partitioning large-vessel traffic based on vessel speed dynamics and their influencing mechanisms. First, speed fluctuations and oscillations are identified using a sliding-window approach and continuous wavelet transform. Bayesian networks, Shapley values, and conditional entropy are then integrated to investigate the factors associated with recurrent speed oscillations. Multi-vessel and multi-voyage oscillation events are spatially aggregated to identify key traffic areas and establish direction-specific traffic zones. Historical trajectories are further used to characterize each zone in terms of vessel speed, actual sailing time, and background-traffic equivalent capacity utilization. A case study on the Jiangsu reach of the Yangtze River produces 49 traffic zones, including 25 key traffic zones and 24 normal navigation zones. The resulting zones provide refined spatial inputs for traffic safety analysis, traffic organization, and large-vessel port entry and departure scheduling.