Research On Passenger Flow Prediction Of Rail Transit Based On AFC Big Data

Urban rail transit is a significant measure to solve road traffic congestion in big cities due to its convenience,comfort,safety,and large traffic volume.What’s more it has become an important part of the urban transportation system.2021 is the first year of the 14 th FiveYear Plan.The plan clearly points out the long-term goals and the suggestions for promoting the construction of infrastructure in a coordinated manner and speeding up the network of urban agglomerations and metropolitan areas.According to the recommendations,the direction of rail transit construction in major cities has also begun to move in the direction of line gridding and technology advancement.Therefore,modeling and forecasting the historical passenger flow of the existing lines can provide the transportation management department with accurating the passenger flow information,which is suitable for them to make timely changes to the train operation plan.Further more,it can also provide strategy to the station worker to handle the passenger flow in emergency.The accurating passenger flow analysis and prediction make a big difference to improving the comprehensive management and the service level of rail transit.Focusing on the AFC analysis and the short-term prediction of urban rail transit,this thesis mainly carries out the following work.(1)We preprocess the original AFC rail transit big data and build a rail transit big data distributed processing platform.And then clean and count the original AFC data.In order to take the feature clearly,we spare no effect to reduce the wrong data.(2)According to the attributes of the land around the Chongqing line site,we select the most representative six types of sites,analyzing their temporal and spatial distribution.We draw their time series diagrams.In order to explore the internal changes of passenger flow,the wavelet analysis is introduced into the track passenger flow sequence.(3)Due to the characteristics of non-linear and high volatility of short-term inbound passenger flow of Chongqing Rail Transit,we combine the STL,ARIMA and LSTM to predict the passenger flow.The STL is a tool to split the flow in some pieces to improving the prediction effect.Based on the STL,we construct ARIMA and LSTM to predict the data splited by the STL.(4)A case analysis was carried out based on the historical data of Chongqing North Railway Station North Square Station.The experimental results show that,compared with the traditional ARIMA,Prophet,LSTM and SVM models.The STL-ARIMA-LSTM model created in this paper performed well,which can fit the forecasting requirements of Chongqing rail transit inbound passenger flow,due to its low MAPE,RMSE and MAE.