A Study On Hotspot Identification And Influencing Factors Of Bus Accidents Based On The Perspective Of Temporal And Spatial Characteristics

The bus is one of the most important means of transport in Chinese cities.Due to the large passenger capacity and large volume,accident often causes serious consequences.In operation management of bus,on-the-spot inspection and video spot check are often used to discover risky driving behaviors.The method of random inspection needs a lot of time and human cost.There is also a lack of accurate understanding of the areas with high incidence.Therefore,it is necessary to identify and analyze the influencing factors in areas with high incidence of bus accidents,which can improve the pertinence of risk areas management and the efficiency of quick emergency response to accidents.In order to analyze the clustering characteristics of bus accidents under the dual constraints of time and space,this paper obtains the accident and violation data from Xi’an Bus Management Company System.Firstly,DBSCAN clustering was used to analyze the spatial distribution characteristics of accident hotspots.Combined with kernel density estimation,the clustering degree was quantified.Secondly,extend the spatial analysis to the space-time dimension,ST-DBSCAN algorithm was used to extract the number and distribution of spatio-temporal hotspots within 24 hours.Spatio-temporal cube was created through aggregating point data to visualize the accident data.Getis-Ord Gi* spatial autocorrelation analysis was used to identify the accident hot and cold spots.The MannKendall time series test was conducted to evaluate the variation trend of hot and cold spots in the time dimension and classified the different patterns of change.In addition,in order to explore the spatial-temporal distribution of bus violations and spatial relationship with accidents,the study also identified the spatio-temporal hotspot of violation events.18 independent variables were considered from five perspectives to establish an analysis framework for the causes of accident hotspots.,such as demographic characteristics,land use pattern,regional road network characteristics,traffic infrastructure and violation location.The logistic regression model was used to analyze the internal influence of risk factors on the formation of accident hotspots.Finally,in order to identify the spatial differentiation and evolution of variables at different times,compared with ordinary least squares model(OLS)and geographically weighted regression model(GWR),spatio-temporal geographically weighted regression(GTWR)model was used.The result of research shows that:(1)ST-DBSCAN algorithm can identify the spatiotemporal location of the accident and violation hotspots.The hotspots distributed in the central area of the city have the widest spatial coverage and longest time span.Getis-Ord Gi*spatial autocorrelation analysis and Mann-Kendall time series test show that the number of accidents and violations has a statistically significant increasing trend.111 accident hotspots were identified,and 166 violation hotspots were identified.The spatial distributions of violations and accidents are similar to some extent.(2)Logistic regression result shows that population density,catering density,landscape density,financial density,house density,bus line density and pedestrian road density have a strong positive correlation with accident hotspots,but education density and subway line density have a negative relationship with accident hotspots.Violation event density has a positive effect on accidents,which indicates that there is a spatial relationship between violation and accident hotspots.(3)Compared with OLS and GWR models,GTWR regression model can fit the influence of independent variables on dependent variables better.The regression coefficients of variables showed that the influence of population density,education density,bus line density,pedestrian road density and violation event density on the accident hotspots has obvious heterogeneity in different time periods.Population density has a positive effect on the accident,the maximum effect intensity was observed at night(16-20).Education density has both positive and negative effects on accidents,which has the most obvious positive effect on accidents in the morning(4-8)and evening(16-20).The characteristics of time dimension are consistent with the commuting peak period;Over time,the positive influence of bus route density on the accident has an increasing trend.The influence degree is the largest in the evening(16-20),and the distribution of high value areas is concentrated.The positive effect of pedestrian road density on accidents was the largest in the morning(8-12).When the high value area covered a wide range,the regression coefficient of violation event density reached the maximum in the evening(16-20),.The above research further reveals the spatio-temporal evolution trend of bus accident and violation hotspots,which provides a basis for accident prevention and control of violation behaviors.It also provides accurate direction and basis for the study of microcosmic accident causes.