X. Li, M. Lang

Pages: 83-94

Abstract
Railway freight volume prediction has great value and significance in railway development planning and allocation of the capacity resources. The combination prediction method based on the support vector machine and NARX neural network is studied in this paper. The impact factors as well as the data processing method, evaluation methods of the prediction performances, prediction idea of the railway freight volume are analyzed. The support vector machine model and NARX neural network model are constructed and by using the Matlab software, the predicted results are calculated. By combining the support vector machine model and NARX neural network model, the linear combination prediction method of railway freight volume is set up and the predicted result is obtained. It is showed that the predicted result of railway freight volume in combination prediction method has higher accuracy than the other two prediction methods and is better than the single prediction method. We choose the combination prediction method as the optimal prediction method in this study.

Keywords: combination prediction method; railway freight volume; support vector machine; NARX neural network