Research On Urban Rail Transit Initial Passenger Flow Forecast Based On Multi-source Data

With the development of urban rail transit in China,more and more cities have entered the"rail transit age".In order to enable the rail transit built by the effort of whole city to serve residents with refined operation and management,passenger forecast is indispensable.At present,the passenger forecast of urban rail transit in China is generally carried out about 5years before the operation of the new line.At the time of line operation,the construction situation and traffic demand around the station may lead or lag behind the basis of passenger forecast.This factor will cause a slight difference between the passenger forecast results and the real passenger flow after the new line operation.Therefore,before the operation of the new urban rail transit line,it is necessary to forecast the initial passenger of new line operation through a low-cost and fast-response mode,so as to provide accurate data for the preparation of the new line.This paper startd with the travel mechanism of urban rail transit travelers,and summarized the influencing factors of urban rail transit passenger into the land use,transportation supply,and land development intensity.After analyzing the current passenger forecast methods,this paper proposed a method to forecast the initial passenger flow of urban rail transit,which integrates spatial econometrics and is based on multi-source data.After that,this paper verified the feasibility and effectiveness of the method by taking Xi’an Rail Transit Line 4 as an example.In this example,this paper firstly collected the POI data,real estate transaction data,bus station data,road network data,taxi operation trajectory data,and urban rail transit operational data of Xi’an in 2017,and integrated these data into a series of indicators of the built environment and travel status around the rail station.Secondly,this paper used exploratory OLS regression and spatial autocorrelation analysis to selecte variables,and determined mandatory travel POI density,non-mandatory travel POI density,residential POI density,pre-owned house rental information number,office rental information number,Taxi travel density and two rail transit network indicators as the independent variables to build forecast model.Thirdly,this paper calibrated the geographically weighted regression model,and predicted the average hourly collector-distributor volume,morning peak hourly collector-distributor volume,and evening peak hourly collector-distributor volume of each station of Xi’an Rail Line 4 in the first year after the operation.Finally,this paper used the 2019 operational statistical data to evaluate the forecast accuracy of the model,compared with the passenger forecast report in the preliminary design stage of the new line,and identified the rail transit stations that should be focused on during the operational preparation.The results showed that the urban rail transit initial passenger forecast method proposed in this paper is highly feasible.The average adjusted R~2 of the geographically weighted regression model was 0.824,and the local fitting residuals were evenly distributed.In the forecast of full-day passenger flow,morning peak passenger flow,and evening peak passenger flow of Xi’an Rail Transit Line 4 in 2019,the relative forecast errors of the entire line were15.41%,23.73%,and 11.21%,and the average absolute forecast errors of each station are 227person/h,575 person/h,and 462 person/h,and the stations with a relative forecast error of(0%,40%)accounted for 74.1%,74.1%,and 77.8%.Based on this,this paper judged that the accuracy of the model can meet the actual work needs,and the forecast results had reference value for the preparation of the new urban rail transit line.