Identification of Vehicle-Pedestrian Collision Hotspots at the Micro-Level Using Network Kernel Density Estimation and Random Forests: A Case Study in Shanghai, China

Abstract

The improvement of pedestrian safety plays a crucial role in developing a safe and friendly walking environments, which can contribute to urban sustainability. A preliminary step in improving pedestrian safety is to identify hazardous road locations for pedestrians. This study proposes a framework for the identification of vehicle-pedestrian collision hot spots by integrating the information about both the likelihood of the occurrence of vehicle-pedestrian collisions and the potential for the reduction in vehicle-pedestrian crashes. First, a vehicle-pedestrian collision density surface was produced via network kernel density estimation. By assigning a threshold value, possible vehicle-pedestrian hot spots were identified. To obtain the potential for vehicle-pedestrian collision reduction, random forests was employed to model the density with a set of variables describing vehicle and pedestrian flows. The potential for crash reduction was then measured as the difference between the observed vehicle-pedestrian crash density and the prediction produced by the random forests models. The final hotspots were determined by excluding those with a crash reduction value of no more than zero. The method was applied to the identification of hazardous road locations for pedestrians in a district in Shanghai, China. The result indicates that the method is useful for decision-making support.

Investigating the Potential of Using POI and Nighttime Light Data to Map Urban Road Safety at the Micro-Level: A Case in Shanghai, China

“…Secondly, the center point of each segment, also known as the reference point (RP), was generated. For each reference point, the density value was calculated as follows [23,39]:…”

“…Commonly, a larger bandwidth may be useful for obtaining hotspots at larger scales, and a smaller bandwidth may be suitable for presenting local effects or hotspot patterns at a smaller scale [39,41,42]. As the former may result in a smooth surface where traffic collision hotspots are prone to mixing with safe neighboring locations, and the latter is likely to produce many tiny isolated hotspots [23], an intermediate value of 250 m was chosen as the bandwidth for this study, to ensure an appropriate density surface.…”

“…Although there are many types of POI data that can be used as variables, introducing excessive variables may cause overfitting of the model [43]. Following previous studies [22][23][24], the variables that are described in Table 3 were chosen for this study. The NTL value of each road segment was taken from the pixel value closest to the reference point.…”

“…Among these explanatory variables, precise traffic exposure data, such as traffic flow and pedestrian flow, are most important but are not easy to obtain. Recently, point of interest (POI) [22][23][24], a type of social sensing data, has been introduced into the crash prediction models due mainly to its easy access and high capability for reflecting characteristics of human activities and the built environment. A typical example is the study by Jia et al [24] that examined the relationship between collisions rate and different types of POI and stated that there were more traffic collisions around banks and hospitals.…”

“…A typical example is the study by Jia et al [24] that examined the relationship between collisions rate and different types of POI and stated that there were more traffic collisions around banks and hospitals. Yao et al [23] have shown that pedestrian collisions are more likely to occur in the vicinity of retail shops [19]. However, previous research mainly focused on vehicle-pedestrian collisions.…”