Container Throughput Forecasting Based On A Hybrid Model Of VMD-ARIMA-HGWO-SVR

Nowadays,global economy has been highly integrated.Marine transportation is becoming increasingly important to global economic and trade chain.As a gathering place for import and export cargos,ports are often accompanied with huge and irreversible investment.Therefore,how to construct the ports more rationally has gradually gained people's great attention.The prediction of port container throughput could help people to determine a reasonable investment scale,optimize the layout,reduce operating costs and make comprehensive overall plan so as to achieve maximum investment returns.Therefore,forecasting the port container throughput accurately has become a research hotspot today.Based on this,this paper proposes a new hybrid decomposition-ensemble model named VMD-ARIMA-HGWO-SVR(VAHS)in order to achieve a higher accuracy of container throughput prediction.Firstly,the original data series is decomposed into several modes(components)by using the variational model decomposition algorithm(VMD).Secondly,the lowfrequency components with less volatility are predicted by autoregressive integrated moving average model(ARIMA).Thirdly,the high-frequency components with greater volatility are predicted by support vector regression models(SVR)which are optimized with a recently proposed swarm intelligence algorithm called hybridizing grey wolf optimization(HGWO).Finally,the prediction results of all modes are ensembled as the final forecasting result.This paper uses the monthly container throughput data of the two largest ports in the world for empirical research.The experimental datasets contain historical data of Shanghai port from January 2001 to May 2016 and Singapore from January 1995 to May 2016.Through the error analysis indicators evaluation of MSE,MAE,MAPE,CDFR and FVD and DM test,the prediction results of the proposed VMD-ARIMA-HGWO-SVR model and other benchmark models are compared.The comparison results indicate that the VMD is more effective than other decomposition methods such as CEEMD and WD,moreover,adopting ARIMA models for prediction of low-frequency components can yield better results than predicting all components by SVR models.The prediction results of VAHS model are more similar to the actual data than other models and achieve the highest prediction accuracy.Thus,the model can be used in practical work to provide reference for the planning,construction,operation,management and scientific development of ports.