Analysis And Prediction Of Urban Macro Traffic Operation Characteristics Based On Big Data Of Ride-Sourcing Trips

The acceleration of urbanization and motorization has led to serious traffic congestion and pollution problems,and the traffic index can provide an objective basis for the governance of urban traffic problems.As the forms of traffic congestion become more and more complex,the evaluation indicators for a single road section and a singlepoint intersection can no longer meet the needs of multi-scale traffic evaluation.There is also an urgent need for indicators that can reflect the macroscopic traffic operation status.Based on this,the city can be analyzed from a macroscopic perspective.Based on the characteristics of traffic operation,it is necessary to promote the control of congestion and prevent congestion.This paper takes this as a starting point,and the research contents and conclusions are as follows:(1)Based on the online car-hailing travel data,a city-level or regional-level traffic index CTSPI is constructed,which enriches the existing traffic operation index system.The algorithm of this indicator does not involve map matching,which can effectively reflect the macroscopic traffic operation state of the city while ensuring simplicity.In addition,CTSPI is sensitive to the date of special events,which is of great significance for the identification of abnormal traffic conditions.(2)Taking the time series composed of CTSPI within the scope of Beijing Sixth Ring Road from December 1,2014 to February 28,2017 as an example,the TBATS model was used to decompose it,and a trend component,two seasonal components and a remaining amount.The seasonal components reveal different degrees of influence of different seasonal factors on CTSPI,among which the annual seasonal effect is greater than the weekly seasonal effect.In addition,compared with STL and classical decomposition methods,the TBATS model filters the seasonal components more fully,and shows the macro traffic trend more clearly,so it is more conducive to observe the trend characteristics in this period.(3)The factors affecting CTSPI were explored from three aspects: travel time,travel restriction policy,and travel environment,and found that it was affected by weekends,holidays,winter and summer vacations,travel restrictions on odd and even numbers,temperature,rainfall,air quality and motor vehicle ownership.Among them,long holidays,restrictions on odd and even numbers,and the number of motor vehicles have the greatest impact.(4)Use a variety of time series methods to forecast the constructed CTSPI time series,including classical exponential smoothing,ARIMA model,and TBATS model that can handle multi-seasonal components.Through comparison,it is found that the prediction results of the TBATS model are better;although the ARIMA model has passed the residual test,the prediction effect is the worst,followed by the exponential smoothing model.