Signal Phase-Driven Traffic State Characterization: A Precision Approach for Urban Intersection Analysis

Data-driven traffic state identification classifies traffic states by mining the operational characteristics of traffic flow, so as to support traffic control and management. Due to the influence of signal phases, it is difficult to accurately classify nonlinear traffic flow. To address this issue, this study proposes a refined traffic state identification method based on intersection phases, aiming at the complex traffic flow characteristics of signalized intersections. First, the collected traffic flow data are segmented and reorganized according to the signal phase scheme. Training samples with “traffic state” labels are obtained using the Fuzzy C-Means (FCM) clustering algorithm optimized by the Rime Optimization Algorithm (RIME), i.e., the RIME-FCM algorithm. Then, the XGBoost model optimized by Bayesian optimization (BO-XGBoost-SHAP) is applied to achieve accurate identification of traffic data. Finally, the effectiveness of the proposed method is verified using real road data. The results show that this method can effectively estimate the traffic flow state of signalized intersections and avoid “false congestion”.