Research On Passenger Flow Forecast Of Urban Rail Transit Based On AFC Data

With the economic development and the accelerating speed of urbanization in China,the influx of population has been increasing.While injecting vitality into the city,it also brings new challenges to the city’s traffic management.It can be seen that in order to Bringing citizens a better travel experience and improving the overall service level of the city,more and more cities have chosen to build rail transit to ease the problem of urban congestion.If there is construction,there must be management.The most prominent direction of urban rail transportation management is passengers.The most striking breakthrough when managing passengers is passenger flow.The improvement of urban rail transportation service levels is inseparable from changes in passenger flow.For a mature urban rail transit operation and management organization,it is very important for a mature urban rail transit operation and management organization to continuously pursue the accuracy of urban rail transit passenger flow forecasting and to explore passenger flow forecasting models with smaller differences.The experimental data and verification data of passenger flow prediction in this paper use AFC statistical data of Shanghai rail transit stations.The concept of station segmentation is proposed for the processing of passenger flow at interchange stations for statistics.The passenger flow data is valid and the experimental basis is reliable.This paper summarizes and summarizes the existing urban rail transit passenger flow forecasting models at home and abroad,and finally locks the time series forecasting model as the basic model for predicting urban rail transit passenger flow,and selects three typical models: ARIMA model,adaptive weights The model is compared with the BP neural network time prediction model as the initial model.In order to refine the research results of the three time series forecasting models,this paper combines the existing urban rail transit station classification basis,and divides the temporal and spatial distribution characteristics of urban rail transit stations through AFC passenger flow into different categories,and shows the three initial models for The results of passenger flow prediction for each type of station draw the conclusion that the typical weight station is most suitable for adaptive stations on weekdays,the BP neural network time series prediction model is best for typical stations on rest days,and the ARIMA model is most suitable for special stations.Introduce the concept of ensemble empirical mode decomposition(EEMD),decompose AFC passenger flow data at different time scales,analyze the impact range of each scale,and divide the passenger flow branches of different time scales obtained by decomposition through reorganization and adjustment of weights,etc.Re-predict and finally get new predicted values.It has been verified that the passenger flow input reconstructed by the EEMD method can effectively predict the change trend of passenger flow,and its accuracy is higher than the initial three time series prediction models,and the model operation is simple,Less time consuming.