Relationship between urban traffic crashes and temporal/meteorological conditions: understanding and predicting the effects

Urban traffic accidents pose significant challenges to public safety and transportation management. Previous studies have revealed that temporal and meteorological factors are the key contributors to accident rate. Besides the inconsistent observations or lack of exploration in some aspects such as snowfall, fog, wind and daily temperatures, it has been shown that these factors are essentially entangled. Furthermore, existing methodologies of analysis or prediction have been limited to relative risk or traditional models. Hence, this study is centered on understanding the detailed correlations between temporal and meteorological factors and accident rate of two types of crashes – moving vehicle and fixed-object crashes using the traffic accident data from Dalian. Further, by incorporating a diverse set of the features, a prediction model leveraging the random forest algorithm is proposed and proved effective in anticipating accident occurrences on the district level. The feature importance analysis has shown that time period and factors such as holiday, temperature and season are most important factors. The rate is higher on working days and in spring, and collisions of both types peak at 6–7 am. When the highest daily temperature is 27 °C and the lowest is 20 °C or -8 °C, the incidence is relatively higher. On the basis, the recommendations are aimed at optimizing transportation systems, mitigating accident risks, and enhancing public safety in urban environments.