Identifying The Factors Affecting The Injury Severity Of Pedestrians And E-bike Crashes:a Data-mining And Statistics Combined Method

Pedestrian and bicyclist’s safety is a growing concern in the world,especially with the growing popularity of e-bike users and the number of people who choose to walk.These road users are considered the most vulnerable to sustain serious injuries since they share the road with automobiles.That is why it is important to identify the factors that influence the severity of bicyclists and pedestrian injuries in crash accidents.This study uses data from the police reported crash data of Hunan province,China,from three consecutive years,2014 to 2016.It will give a deeper understanding on what factors affect pedestrian and e-bike crash accidents,and also give some suggestions on the measures needed to reduce or eliminate these crashes.An integrated method of K-Means clustering analysis and Multinomial Logistic Regression model(MNLR)were integrated to analyze the contributing factors affecting the injury severity and explore the impact of such factors on three different severity levels of pedestrian crashes.A K-means cluster analysis was applied to segment the data into homogeneous groups.Moreover,multinomial logistic regression models were built for the whole dataset and for each cluster to analyze the crash data,with the possible outcomes being fatalities,severe injury and minor injury.Furthermore,a combined method of integration,a tree classification and logistic regression model,was developed to examine the underlying risk factors of severe injury for different groups of e-bike riders.E-bike riders were segmented into five different groups based on the tree classification analysis;nonoccupational riders aged over 55 in urban regions are associated with higher likelihood of severe injury among the five groups.The results suggested that the significant factors affecting the injury severity of pedestrian crashes included infractions committed by the driver(fatigue,drunk driving,driving against the traffic light,reverse driving),infractions committed by pedestrians(walking on the driveway),visibility,pedestrian age(above 65),weather conditions and road type.The results of the logistical analysis for each group show that several risk factors such as high-speed roads have commonly significant effects on injury for different groups,whereas major factors only have significant effects for specific groups.The conclusions drawn from this study provide policy implications to alleviate the crash injury for different e-bike riders groups.