Analysis Of Pedestrian Injury Severity By Considering Data Heterogeneity

Slow traffic is an important part of modern urban traffic system.Indeed,by considering the growing levels of congestion and higher parking costs,slow traffic,as an irreplaceable mode of transportation,has been paid more and more attention.As the main body of practicing slow traffic,pedestrians are more and more exposed to the urban transportation system.Hence,the conflicts between pedestrians and other modes of transportation are increasing and the pedestrian safety problems are becoming more severe.Among the pedestrian casualties,intersections accounts for major proportions instead of mid-block segments,thus it is necessary to investigate the pedestrian injury severity at intersections.In this thesis,pedestrian injury severity is concentrated on by considering the unobserved heterogeneity problem,which may lead to model coefficient estimation deviation without being included.The Bayesian quantile regression model and geographically and temporally weighted regression model are proposed to deal with the spatial and temporal heterogeneity problem respectively,to identify the influencing factors of pedestrian injury severity at signalized intersections.The effects of various influencing factors on pedestrian injury severity are revealed and improvement measures are presented from an empirical perspective.Firstly,by focusing on the temporal heterogeneity problem resulting from data collected from different time,the Bayesian binary quantile regression model and Bayesian ordinal quantile regression model are proposed to analyze the pedestrian injury severity at signalized intersections in city A,and the results of the two models are compared.The results show that it is feasible to take the pedestrian injury severity as a binary or ordinal variable and Bayesian binary quantile regression model is better than Bayesian ordinal quantile regression model in terms of goodness-of-fit.After that,by including the temporal and spatial heterogeneity problem resulting from data collected from different places and time,the geographically and temporally weighted regression model are put forward to analyze the pedestrian injury severity at signalized intersections in city B.The significant test on the influencing factors of pedestrian injury severity are conducted,and the spatio-temporal variation of different influencing factors are revealed.The results display the spatio-temporal non-stationary in traffic accident data.Finally,the discussions are made and conclusions are reached,and then further study directions are orientated.