Research On Railway Passenger Flow Forecast Based On Time Series

With the rapid development of China’s social economy and the continuous improvement of people’s living standards,people have further requirements for the efficiency and experience of public transport travel.High-speed rail is now a popular means of transportation for public travel.In order to improve the railway department’s passenger service management capacity and reverse the situation of insufficient line revenue in some regions,the railway department needs to formulate by understanding the daily railway passenger traffic,the seasonal and seasonal changes index of peak and off-peak periods,and the specific conditions of cold and hot lines Reasonable train operation plan,among which,the research of railway passenger flow prediction is the key to promote railway passenger transportation service capacity and rational allocation of resources.This paper selects the daily passenger flow data of a high-speed railway passenger dedicated line from January 2015 to March 2016,uses feature engineering to preprocess and analyze the correlation of the original data,and determines the time series that can be used in the prediction model data set.Through the analysis of the characteristics of railway passenger flow and the study of time series forecasting methods,three algorithms of ARIMA,LSTM and Prophet are selected to predict and study railway passenger flow.Based on traditional time series forecasting methods,ARIMA and LSTM models are constructed to predict and analyze the future passenger flow of the railway respectively.Considering that railway passenger flow is simultaneously affected by seasonal cycles,emergencies,and holidays,a new time series forecasting algorithm Prophet is used in this paper,and a Prophet forecasting model is established based on this principle.By comparing and analyzing the prediction errors of different single-item models,it is found that the prediction accuracy of single-item models is insufficient,and it is proposed to build combined models ARIMA-Prophet and LSTM-Prophet on the basis of single-item model prediction.Through the use of visualization methods and model prediction evaluation indicators to make a comprehensive comparative analysis of the prediction errors between different models.The results show that the improved combined model’s prediction performance is significantly higher than that of the single-item model,which can more accurately predict the future passenger flow of the railway.