Research And Implementation On Short-term Passenger Flow Prediction Of Urban Rail Transit Based On Deep Learning

The short-term passenger flow prediction of urban rail transit is an important decision-making basis for rail transit operation enterprises to formulate driving plans and passenger transport organization plans.At the same time,it is also an important part of developing intelligent transportation services,intelligent passenger services and building digital and intelligent urban rail transit.It is of great significance to help the digital transformation of rail transit and build intelligent rail transit.This thesis takes the short-term passenger flow of Shijiazhuang subway as the research object.Based on the analysis of the spatial-temporal distribution characteristics of the passenger flow of Shijiazhuang subway network,this thesis constructs the line level short-term passenger flow prediction model based on the improved particle swarm optimization short-term memory network and the network level short-term passenger flow prediction model based on convolution neural network,and studies the short-term passenger flow prediction of rail transit lines and networks respectively.Based on the above research,we design and implement the short-term passenger flow prediction system of rail transit.The major contents are as follows:(1)Analysis of passenger flow characteristics of rail transit.Based on the original transaction data of Shijiazhuang subway AFC system,the basic data set is obtained after data preprocessing.From the perspective of different dates,different periods and different stations,this thesis analyzes the spatial-temporal distribution characteristics of the passenger flow in Shijiazhuang subway network.(2)Short term passenger flow prediction of rail transit line level based on IPSOLSTM.Based on the long short-term memory neural network,the short-term passenger flow prediction model of rail transit line level is established.Firstly,the inertia weight and learning factor in the standard particle swarm optimization algorithm are improved,and then the improved particle swarm optimization algorithm is used to optimize the parameters of LSTM neural network,and the rail transit line level short-term passenger flow prediction model is established.Compared with the traditional SVM model and the basic LSTM model,the prediction accuracy is improved.(3)Short term passenger flow prediction of rail transit network level based on CNN.Based on convolution neural network,the short-term passenger flow prediction model of rail transit network level is established.Firstly,the passenger flow data is processed into a two-dimensional matrix with time as the row and passenger flow at each station as the column,and then CNN neural network models with different depth structures are constructed and trained for fine parameter adjustment to obtain the optimal structure and parameters of the model,and the model construction is completed to realize the synchronous prediction of short-term passenger flow at all stations of rail transit network.(4)System of rail transit short-term passenger flow prediction.The short-term passenger flow prediction system of rail transit is designed and implemented by using Python language,MySQL database and Vue front-end framework.The system realizes the functions of short-term passenger flow prediction and passenger flow data statistical analysis of rail transit lines and stations.