Revision And Comparison Of Short-term Forecasting Models For Railway Passenger Traffic Volume In China

As a major mode of transportation,the railway has brought great convenience to people's travel.In order to improve the accuracy of short-term forecasting of railway passenger traffic in China,the following work has been done:The first chapter first introduces the research status of passenger traffic forecasting,combs the commonly used quantitative forecasting model and its modified model,and then compares the advantages and disadvantages of the model.Finally,the main work of this paper is expounded.The second chapter starts from the annual data and takes the multi-collinearity problem between variables as the starting point.It mainly introduces five commonly used models that can be used to correct the problem.These models include stepwise regression,ridge regression,adaptive LASSO regression,principal component regression,and partial least squares regression.In order to more fully extract the seasonal factors and trend factors in the sequence,when analyzing the quarterly data of railway passenger traffic,the time series decomposition method,the seasonal cycle regression model and the seasonal multiple regression model are considered for correction.On the basis of the model correction,the model is constructed by combining the quasireciprocal method and the entropy weight method to further improve the stability and prediction performance of the model.The third chapter makes an empirical analysis of the annual data prediction model of China's railway passenger traffic.The empirical analysis shows that the fitting errors of these five models are all controlled within 3%,that is,the fitting effect of each model is better.At the same time,the three models with the extrapolation predictive power ranking are selected: ridge regression,partial least squares regression and principal component regression.Then,the three models are combined and predicted by the quasi-reciprocal method and the entropy weight method.It is found that the combined model based on ridge regression and principal component regression has the smallest error and its error drops to about 1/3.Then,the obtained prediction results are compared with the common models.The results show that the prediction error of the model constructed in this paper is about 1/5 of the grey regression combination model and 1/4of the quadratic exponential smoothing model.The fourth chapter makes an empirical analysis of the quarterly data forecasting model of China's railway passenger traffic.When constructing the quarterly data prediction model,the prediction models are first corrected accordingly,the error of the corrected model decreased to 1/6,1/13 and 1/5 of the original model,respectively.Then the modified model is combined and predicted,and the combined combined prediction results are compared with the common time series model.The comparison results show that the combined model has better prediction performance,and the prediction error is about 1/9 of the SARIMA model and 1/5 of the Holt-Winters addition model.