Short Term Passenger Volume Forecast Of Urban Rail Transit OD Section Based On LSTM

AFC(Automatic Fare Collection)system is widely used in urban rail transit to record passengers’ travel data,but AFC can not record passengers’ travel path.As urban rail transit enters into network operation stage,path matching and sectional passenger flow calculation of rail network become a difficult problem.Accurate prediction of short-term passenger flow between stations and getting the distribution of short-term passenger flow in future urban rail transit network is of great significance for urban rail transit network planning,management,operation and improving passenger travel experience.Firstly,the structure of passengers’ travel time is studied.According to the density peak clustering results,the number of effective paths between stations is analyzed.Then,the K-means clustering algorithm is used to divide the travel time of multi-path stations into time sets and match the travel paths of each time set.Secondly,combined with the urban rail train timetable,the matching method of passenger number without transfer and with transfer path is proposed,and the passenger number of each train between corresponding stations is calculated,so as to count the short-term passenger flow between stations.Thirdly,the short-term passenger flow prediction model of urban rail transit station section based on LSTM(long short term memory)neural network is proposed.The short-term passenger flow of future station section is predicted by stacked LSTM neural network and historical short-term passenger flow of station section.Finally,taking the AFC card data of W city rail transit as an example,the data cleaning of passenger AFC card data and the mining of its internal information are realized by programming.The validity and prediction accuracy of the model are verified from four aspects: travel time cluster distribution,path matching,cross-section passenger flow calculation and short-term station section passenger flow prediction.