Spatiotemporal Influence Analysis Through Traffic Speed Pattern Analysis Using Spatial Classification

This study introduces a method for classifying traffic flow segments on expressways to estimate impact zones in merging/diverging sections and accident-prone sites. I propose a spatiotemporal dynamic segmentation approach that enables real-time identification of traffic hazard sections, reflecting changes in traffic flow, as opposed to traditional traffic analysis based on predefined segments in a node–link network. This methodology uses high-resolution vehicle trajectory data to precisely identify unstable and low-speed traffic sections. Using the geohash algorithm, the area is hierarchically segmented based on the standard deviation of speed in general traffic flow, facilitating the identification of unstable traffic flow patterns. For eight expressway routes, traffic flow was categorized into stable or minimum-size spaces. From a total of 1207 segments, 943 unstable flow segments were identified. The impact zones of the merging and diverging sections on Expressway 50 were analyzed using the results of spatial segmentation. Furthermore, by comparing traffic data before and after accidents, I assessed the short- and long-term effects of accidents on traffic flow. The proposed methodology provides precise data essential for reducing the likelihood of traffic accidents and for predicting post-accident congestion and duration. The patterns of such accident impact zones can contribute to preventing secondary accidents by providing advance information to following vehicles through various communication methods, including those involving autonomous vehicles. This enables effective traffic management strategies and rapid responses to accidents.