The topology of urban road networks and traffic conditions are complex, are more sporadic and are influenced by the driving environment, which poses significant challenges for evaluating traffic flow states. This study focuses on the spatiotemporal correlation characteristics of urban transportation systems and the interaction between individual and collective behaviors. Through traffic flow data mining, it explores the dynamic evolution patterns of traffic states in urban road networks. A Tri-State Traffic Factor Network (TS-TFN) model is proposed to observe multidimensional states at microscopic (individual vehicles), mesoscopic (road segments) and macroscopic (road network) levels. This enables precise modelling and in-depth characterization of urban traffic flow. Accurate state estimation is achieved by comparing results from multiple clustering algorithms, such as GMM and Agglomerative. Based on the dynamic evolution mechanism of traffic states, the tri-state characteristics of the road segments in the constructed model are especially well-suited to research into urban transportation, with a view to elucidating the intrinsic mechanisms of multi-level traffic dynamics.