Severity of Pedestrians in Pedestrian - Bus Crashes: An Investigation of Pedestrian, Driver and Environmental Characteristics Using Random Forest Approach

Bus- pedestrian crashes typically result in more severe injuries and deaths than any other type of crashes due to physical characteristics such as heavyweight, large size and maneuvering restrictions. The statistical data report by the Ministry of Road Transport and Highways (MoRTH, 2017) highlights that bus crashes alone account for about 6.9% of the total crashes occurring within the country. 10,651 (7.2%) persons have been killed, and 44,330(9.4%) persons have been injured in bus-involved collisions during 2017. The purpose of this research is to investigate the factors that significantly contribute to the severity of pedestrian injuries resulting from pedestrian- bus crashes using a random forest approach. Contributory factors, including the driver, pedestrian, and environmental characteristics, are investigated and discussed. The crash dataset for the present study was prepared from the police-reported pedestrian- bus crashes for the past nine years that occurred within Tamilnadu. The research team retrieved all the single-vehicle out-of-control four-wheeler crashes for the period between 2009 and 2017. Random Forest method is an ensemble method for classification problems that is a collection of decision trees. It aggregates all the predictions made by the decision trees into one final prediction. The complete dataset (11735) was divided into two separate datasets: the training dataset (9388) for the development of the model and the testing dataset (2347) for the performance evaluation of the model. The most significant variables in RF were found to be the number of lanes (both single and two-lane), presence of median separators, crashes occurrence at the intersection was unknown and pedestrians aged above 55 years. Based on the findings of the above results, targeted countermeasures may be designed in light of the injury severity of the pedestrians in pedestrian- bus collisions.