Prediction And Factors Analysis Of Railway Freight Volumes

As an important representation of railway transportation capacity, it is really important to grasp the development trend of railway freight volumes. It can provide valuable advice of marketing plan for railway transportation enterprise, provide accordance for department of railway transportation to make policies and offer methods for investments of railway construction.The aim of the paper is to find most valuable model for railway freight volumes and predict the future. In addition, the paper finds key factors affecting railway freight volumes and puts up politic advice to control railway freight volumes.First, the paper uses railway freight volume data from 1990 to 2014 to build several time series models and chooses the model that has smallest average relative error as the best time series model. The best time series model is the quartic polynomial time series model of railway freight volumes. The paper uses this model to predict the future, the railway freight volumes in the next five years are 406,8335kt,407,7049kt, 404,4083kt,396,420.8kt and 383,200.8kt. The railway freight volumes will decrease in the future.To change the trend of the railway freight volumes, the paper uses grey correlation analysis to find key factors for railway freight volumes and uses these factors to build multivariate linear regression model and neural network model. The average relative error of multivariate linear regression model is 0.7398%. The key factors that are significant are national gross income of finance, production of steels, production of kerosene, highway freight volumes and waterway freight volumes. The increase of national gross income of finance, production of steels and production of kerosene will result in the increase in railway freight volumes, while the decrease of highway freight volumes and waterway freight volumes will increase in railway freight volumes. The paper uses these characteristics to give political advice. Compare time series models, neural network model and multivariate linear regression model, the paper finds that the most valuable model is multivariate linear regression model.The innovation point of the paper is that it chooses best time series model to predict the future of railway freight volumes and uses multivariate linear regression model based on grey correlation analysis to find key factors and finally gives political advice to increase railway freight volumes.