To realize the coordinated control of multiple intersections in a large-scale urban road network, it is necessary to exploit a method to study critical intersections’ global correlation characteristics. First, the Real-coded Accelerating Genetic Algorithm Projection Pursuit Classification (RAGA-PPC) is used to solve and evaluate the intersections critical degree. Second, the k-means clustering algorithm is applied to intersections critical degree to divide the critical intersections and non-critical intersections. Third, a new approach of calculating the critical intersections’ correlation degree based on the improved FP-Growth algorithm, which can mine frequent itemsets of intersections from vehicle trajectories, is applied to analyze the global correlation characteristics of the selected critical intersection. Finally, Didi GAIA Dataset is selected for critical intersections identification and correlation analysis. The experimental results show that: the correlation between critical intersections tends to decrease with distance globally and is time-varying.

CICTP