Port traffic flow prediction with RF-ARIMA-BLSTM

To improve the rate of vessel entry and departure at ports, enhance the accuracy of ship traffic flow prediction, and meet the future development needs of ports, a prediction method based on RF-ARIMA-BLSTM is proposed. The ARIMA model is combined separately with ISTM and BLSTM neural network models incorporating Random Forest（RF） factor selection for prediction and comparative analysis. This approach is applied to forecasting the monthly total vessel arrivals and departures at Qingdao Port（2016—2019） and Dalian Dayao Bay Port（2011—2020）. The results show that the RF-ARIMA-BLSTM method consistently achieves the highest prediction accuracy. Compared with four other prediction methods, its evaluation metrics-including root-mean-square error（RMSE）, mean absolute error（MAE）, and mean absolute percentage error（MAPE）-are all the lowest, verifying the effectiveness of this method. The proposed ship traffic flow prediction approach is expected to provide decision-making guidance for future port development and planning.