| Railway freight is an important part of my country’s transportation industry.In recent years,with the continuous changes in the international economic situation and the introduction of relevant policies in my country,the railway freight market has developed rapidly.The relevant railway management departments need to keep abreast of the freight transportation situation,grasp the development trends of the freight market,make route planning in advance,and reasonably allocate transportation resources.Therefore,accurate forecasting of freight demand is conducive to the planning of future development of the railway sector,and it is particularly important to develop a scientific and effective railway freight demand forecasting method,which has certain guiding and reference significance for the work of the railway sector.At present,domestic and foreign scholars have relatively comprehensive research on railway freight forecasting technology.Generally speaking,they can be divided into two categories.One is to make predictions by analyzing historical freight data without considering influencing factors,and the other is to predict by analyzing the relationship between the sequence to be predicted and related influencing factors.This paper conducts research on the problems of these two types of forecasting methods.The main contents of the work are as follows:(1)A freight demand forecast model based on unit phase space reconstruction is proposed.The method of predicting the law of future development based on historical data is subjective,and the prediction accuracy is limited in the face of fluctuating freight series.Therefore,chaos theory is introduced to reconstruct the unit time series into a multi-dimensional phase space,and the C-C method for solving reconstruction parameters is improved,the input structure is determined and the nonlinear model is used for prediction.At the same time,considering the limitations of using a single prediction model,the idea of decomposition and integration is introduced to reduce the prediction error,and the original freight sequence is decomposed to obtain low-frequency and high-frequency components.The linear model is used to predict the low-frequency components,and the nonlinear model is used to predict the high-frequency components.Finally,the prediction results of each decomposition component are integrated.(2)A freight demand forecasting model based on multi-phase space reconstruction is proposed.The method of forecasting freight demand by analyzing influencing factors is usually to use common models to model multi-variables to approximate the original freight system,but this does not fully excavate the true nature of the freight system and affects the prediction accuracy.First,it analyzes and summarizes the influencing factors of freight demand,and selects key factors to establish an index system.Secondly,the reconstruction parameters of the sequence to be predicted and the influencing factor sequence are solved,the chaotic characteristics of each sequence are judged,and the multivariate phase space is constructed.Finally,the combination of the multivariate phase space matrix and the neural network is used to predict,which provides a new idea for the existing freight forecasting methods.This paper verifies the effectiveness of the two forecasting models on the freight container data set and compares them with other forecasting methods.The experiment proves that the two forecasting methods proposed have higher accuracy. |
