Citation

Pour AT, Moridpour S, Rajabifard A, Tay R. Road Transp. Res. 2017; 26(1): 4-20.

Copyright

(Copyright © 2017, Australian Road Research Board)

DOI

unavailable

PMID

unavailable

Abstract

About 1100 vehicle-pedestrian crashes occur in Melbourne metropolitan area every year. Identifying the temporal and spatial patterns of pedestrian injuries is essential to enhance the safety of these vulnerable road users. In this paper, Decision Tree (DT) and interactive DT are applied to identify the influence of temporal, spatial and personal characteristics on vehicle-pedestrian crash severity. DT is a simple but powerful form of data analyses using machine learning technique. Result of DT indicates that time of crash is the most significant variable in classifying and predicting the severity of vehicle-pedestrian crashes in Melbourne metropolitan area. According to this model, accidents occurring between 7:00 p.m. and 6:00 a.m. are more severe than other times. Moreover, spatial correlation shows that there are positive correlation between time and location of crashes. Kernel Density Estimation (KDE) is applied to explore the spatial distribution of vehiclepedestrian crashes. KDE results show that most vehicle-pedestrian crashes between 7:00 p.m. and 6:00 a.m. occur around hotels, clubs and bars. Safety measures should be applied around these areas to assist in preventing and reducing the severity of vehicle-pedestrian crashes.

Language: en