Prediction models based on multivariate statistical methods and their applications for predicting railway freight volume

Abstract

Four prediction models based on the multivariate statistical methods are constructed in this work and they are successfully applied in predicting the Railway Freight Volume (RFV). RFV directly reflects the regional economic states such as production improvement and economic restructuring. Accurately predicting the RFV is of great use in production planning, decision making, labor allocating, etc. In this work, based on the multivariate statistical methods, i.e. ordinary least squares regression (OLSR), principal component regression (PCR), partial least squares regression (PLSR), and modified partial least squares regression (MPLSR), four RFV prediction models are constructed and the detailed comparison is made by implementing them on a practical dataset. From the simulation results, the conclusion can be derived that the MPLSR based prediction model outperforms the other three models.