Research On Subway Passenger Flow Forecast Based On Improved BP Neural Network

With the development of economy and the rapid increase of people flow,the number of road vehicles is also increasing.In order to reduce the traffic pressure on the road,the role of subway becomes more and more important.With the gradual increase of subway lines and the gradual completion of subway lines,the temporal and spatial distribution of subway passenger flow has changed to varying degrees.Accurate and complete prediction of the short-term passenger flow of subway passenger flow is of great help to the formulation of subway operation plan,the reduction of accidents caused by personnel flow,and the formulation of subway traffic scheduling plan.Therefore,the subway passenger flow forecast is particularly important.Based on the passenger flow of Nanchang Metro Line 2 as the data support,this paper carries out the relevant research on the passenger flow forecast of Nanchang Metro Line 2(1)The theoretical and experimental data of passenger flow of Nanchang Metro Line from December 31,2019 are collected and analyzed.According to the relationship between the data,the missing data and error data are processed,and a BP neural network prediction model is constructed about the relationship between subway passenger flow and historical weather data,historical time data and historical passenger flow data.(2)After training the training data with BP neural network prediction model,the passenger flow of Nanchang Metro Line 2 in July and August 2020 is predicted,and compared with the actual passenger flow.It is found that the prediction quantity of BP neural network has certain accuracy,and a combined model is designed for passenger flow prediction.(3)The principle,advantages and disadvantages of genetic algorithm are studied.The chromosome coding,the selection of the appropriate adaptive function and the selection of the three operations in the genetic algorithm are carried out.The operation method which is more in line with the experimental requirements of this paper is selected.By modifying the genetic algorithm and improving the selection of weights and thresholds in BP neural network model,a combination model is established by combining the improved genetic algorithm and BP neural network.(4)The BP neural network prediction model,the BP neural network prediction model based on genetic algorithm and the improved combination model are predicted at the same time.The results show that the accuracy of the improved combined model is 4%~ 6% higher than that of the traditional BP neural network.