| With the steady development of the Chinese social economy and the implementation of the strategy of strengthening the country in transportation,the high-speed railway network has continued to expand.Accurately predicting the amount of high-speed railway passenger tickets purchased during the pre-sale period is conducive to grasping the changes in passenger demand.This is of great significance for the railway sector to optimize ticket allocation and improve the service level of passenger transportation.Recently,deep learning technology has shown powerful big data analysis performed in the field of intelligent transportation.In the face of the huge high-speed railway passenger transportation market,deep learning technology can be used to integrate a variety of passenger ticketing factors to grasp the changes in passenger ticketing volume for each day of the ticket pre-sale period in real time.Thus,the construction of an efficient passenger ticket pre-sale period ticket purchase prediction model for the railway sector has become an urgent problem to be solved.This paper firstly expounds the basic concept and data processing process of predicting the ticket purchase volume during the pre-sale period of high-speed railway passenger tickets.Then,it analyzes the features of ticket purchases from three aspects: the distribution features of ticket purchase data in the historical continuous departure period,the date corresponding to the departure date of the ticket,and the attributes of holidays.Based on the above analysis of the features which are included in the pre-sale period of high-speed railway passenger tickets,this paper uses convolutional neural network(CNN)model,long short-term memory(LSTM)neural network model,and other neural networks to build a deep learning CNN-LSTM combined prediction model for the pre-sale period of high-speed railway tickets based on multifeature fusion.In the case analysis,the historical ticket purchase data of passengers on the Shanghai-Kunming high-speed railway from Changsha South to Nanchang West and from Changsha South to Hangzhou East provided by the railway department are used.The autoregressive integrated moving average(ARIMA)model,BP neural network,convolutional neural network(CNN),and long short-term memory(LSTM)neural network models are used as comparative prediction models.The root mean square error,mean absolute error,and mean absolute percentage error are selected as the evaluation indexes of the prediction model.In this study,it’s shown that Chinese high-speed railway passengers often concentrate on purchasing tickets close to the departure date of the tickets,and the data on the number of tickets purchased by high-speed railway passengers on each day of the pre-sale period corresponding to the historical consecutive departure periods fluctuate more regularly,while the number of tickets purchased by high-speed railway passengers during the pre-sale period of special periods such as holidays fluctuate more significantly.When using the combined CNN-LSTM prediction model with multi-feature fusion for deep learning of high-speed railway pre-sale ticket purchase volume,the root mean square error between Changsha South to Nanchang West OD is 24.0,the mean absolute error is 11.7,and the mean absolute percentage error is 1.1%,and the root mean square error between Changsha South to Hangzhou East OD is 13.7,the mean absolute error is8.1,and the mean absolute percentage error is 1.7%.The error accuracy of the mean absolute percentage error of the combined prediction model is better than that of the comparative prediction model.The combined model has better predictive performance compared to the comparative model and can use historical data to predict future ticket purchases during the presale period.This provides a certain theoretical reference for the dynamic adjustment of ticket amounts in the railway passenger transportation market,thereby improving the quality of passenger transportation services. |
PDF Full Text Request