Hotspot Segments Identification of Collisions Between Electric Vehicles and Pedestrians/Non-Motor Vehicles

To explore in-depth the spatiotemporal distribution characteristics of collisions between electric vehicles and vulnerable road users, this paper proposes a method for identifying spatiotemporal hotspot segments of collisions between electric vehicles and pedestrians/non-motor vehicles. First, based on the collision data involving electric vehicles and pedestrians/non-motor vehicles, the analytic hierarchy process is employed to determine the weights of the influencing factors, and a weighted network kernel density estimation method is employed to reveal the spatial clustering of traffic accidents. On this basis, the density peaks clustering (DPC) algorithm is utilized as a spatial clustering model for accidents, and a spatiotemporal-DBSCAN (ST-DBSCAN) model is constructed to incorporate the temporal dimension, thereby accurately characterizing the spatiotemporal distribution characteristics of collision accidents between electric vehicles and pedestrians/non-motor vehicles. Finally, an empirical study is conducted using the electric vehicle accident data from a city over 11 consecutive months. The results indicate that the collisions between electric vehicles and pedestrians/non-motor vehicles exhibit 3 temporal peaks, differing from the traditional bimodal characteristic observed in traffic accidents, while spatially demonstrating localized clustering characteristics. For identifying spatial hotspot segments, as compared with the optimal values of DBSCAN, OPTICS and Mean Shift algorithms, DPC algorithm shows improvements of 42.9%, 74.5% and 11.1% in terms of silhouette coefficient, Davies-Bouldin index (DBI) and Calinski-Harabasz Index (CHI), respectively. For identifying spatiotemporal hotspot segments, under similar DBI conditions, ST-DBSCAN algorithm achieves silhouette coefficient and CHI values that are 2.25 times and 57.3% higher, respectively, than the optimal values of ST-OPTICS, ST-DPC and ST-Mean Shift algorithms.