China‐Europe Railway Express Freight Volume Forecasting With a Novel SARIMAX‐BNN Model

This study proposes a hybrid SARIMAX‐BNN model to forecast China‐Europe Railway Express (CR Express) freight volume, addressing challenges such as nonlinearity, temporal dependence, and high volatility. Spearman rank correlation is used to select key influencing factors to optimise SARIMAX inputs. The SARIMAX model captures trend and residuals, while the BNN learns nonlinear residual patterns. Final predictions are obtained by combining both models. Using annual and monthly freight data from Xi'an CR Express, the proposed model achieves an MAE of 0.16 (10k tonnes) and a MAPE of 0.90% on the annual dataset and an MAE of 14.53 TEU with a MAPE of 0.05% on the monthly dataset, outperforming all baseline methods. This approach improves forecasting accuracy and supports better operational decisions.