Research Of Rail Traffic Flow Forecasting Based On Delay Nonlinear Autoregressive The Neural Network

With the rapid development of the city,the conflict between the travel demand of urban population and urban traffic is becoming more and more prominent.Urban rail transit,with its unique advantages such as high speed,high capacity and environmental protection,stands out from a variety of modes of transportation,becoming the main means to solve traffic congestion.Cities have been put into construction,so that the urban rail transit from single line operation to the line network operations,the size and complexity of the upgrade,but also on the management of rail transit network and development challenges.The rapid and accurate passenger flow forecasting is not only the basis for the scientific development of the traffic plan,but also an important basis for real-time adjustment of the operation plan.In this paper,based on the historical passenger flow data of Chongqing rail transit,the characteristics of passenger flow and spatial and temporal distributio n are analyzed.The main work and innovation of this paper are as follows:(1)This paper discusses the characteristics of the network and the characterist ics of passenger flow,and analyzes the spatial and temporal distribution of the p assenger flow in Chongqing.(2)Research the theoretical basis of prediction of passenger flow,analyzes th e research experience and previous common prediction methods,pointed out that the support vector machine has outstanding performance in solving the problem of small sample and high dimension pattern recognition and so on,but the pred iction for the nonlinear data large amount of data,complex still exist limitations.Furthermore,it is pointed out that the nonlinear recurrent neural network has be tter performance in dealing with the nonlinear data,and can make up the limitati on of the support vector machine.(3)In this paper,a passenger flow forecasting model based on support vector regression is constructed.Based on the NAR neural network model of passenge r flow forecasting,in order to improve the prediction accuracy,the NARX neural network with external input is introduced and the prediction model is establishe d.The network structure and parameters of the neural network are designed,and several training algorithms and excitation functions are compared.(4)All lines of Chongqing rail transit(line 1,line two,line three and line si x)in 2014 January,the passenger flow data as the foundation,through data prep rocessing,statistical value data into three prediction model for experiment.Accor ding to the characteristics of passenger flow theory,experimental design and pred iction of short term forecast of passenger flow peak,according to the design of neural network model to commute to school and district based representation of the passenger flow distribution forecast site.(5)In the design of Chongqing rail traffic coordinated development analysis p latform forecast module,the module from the operation indicators,line passenger traffic,line section and distribution fare aspects of Chongqing rail transit system to provide a scientific basis for daily operation.The module provides a variety of commonly used prediction algorithms,but also provides an external interface f or researchers to add new prediction algorithm at any time.