Research On Short-term Passenger Flow Forecast Of Urban Rail Transit Line

With the development of urban rail transit has been flourishing in China,Many cities have ended up the single-line operation mode and stepped into the network operation mode.Short-term passenger flow forecast of urban rail transit(URT)line is the basis of daily management of the URT's operating department,which is highlighted under the condition of network operation.Therefore,based on the historical passenger flow data of URT,time characteristics and the forecasting model of short-term passenger flow of URT line are studied in this thesis.First of all,the short-term passenger flow factors are summarized as five influencing factors of ticket price,networking effect of URT,service level of URT line,seasonal factors and temporary factors to analyze in this thesis.And then,based on the analysis of seasonal variation characteristics and daily variation characteristics of URT lines' passenger flow,this part focuses on the research of variation characteristics of ordinary-day and holiday passenger flow,which lays the foundation for the construction of forecasting model.Then,Support Vector Regression(SVR)is chosen to kernel method of short-term passenger flow forecast according to the nonlinear characteristics of short-term passenger flow and the limitation of sample data.The PSOGA algorithm is proposed to optimize parameters of SVR,which is combined with the advantages of Genetic Algorithm(GA)and Particle Swarm Optimization(PSO),that is,the PSO formula is introduced as the mutation operator of GA.In the view of the obvious seasonal fluctuation characteristics of holiday passenger flow,a PSOGA-SVR model based on Seasonal Exponential Adjustment(SEA),called SEA-PSOGA-SVR model,is presented to forecast holiday passenger flow.Eventually,daily inbound passenger flow data from Nanjing Metro Line 2 in years 2013 to 2016 is taken as a case.The results showed that root mean square error and mean absolute percentage error of PSOGA-SVR model are less than the corresponding errors of GA-SVR model and PSO-SVR model.Therefore,the PSOGA-SVR model can forecast short-term passenger flow greatly with low errors and high prediction accuracy.Besides,the SEA-PSOGA-SVR model is used to forecast holiday passenger flow,and its root mean square error and mean absolute percentage error decrease obviously,which can eliminate the seasonal influence of the holiday original data by using the seasonal index adjustment method to deal with the seasonal characteristics of the original data directly,so it can improve the accuracy of the PSOGA-SVR model forecasting holiday passenger flow.