Short-Term Prediction of Ship Traffic Flow Based on SABO-VMD-HPO-GRU

To enhance the prediction accuracy of ship traffic flow, considering its nonlinearity and non-stationarity, this paper proposes a ship traffic flow prediction model based on SABO-VMD-HPO-GRU. Firstly, the preprocessed ship traffic flow data are decomposed using VMD. To obtain better decomposition results, the SABO algorithm is employed to optimize the number of decomposition modes (K) and the penalty factor (α) in the VMD model before decomposition. Subsequently, the IMF components are input into the GRU network for training. Prior to training, the HPO algorithm is utilized to optimize parameters in the GRU network. Finally, the training results are aggregated and reconstructed to obtain the final prediction results. This model is validated using actual data from the waterway near Ningbo Yongzhou. The experimental results demonstrate that the optimized VMD-GRU model, compared to the unoptimized one, reduces the RMSE, MAE, and MAPE by 42.1%, 42.6%, and 44.5%, respectively. Additionally, the R2 increases by 11.4%. The results show that the model achieves a prediction accuracy of 90.04%, indicating higher precision.