Forecasting Research Of Freight Volume Of Ports And Aviation Based On Secondary Decomposition-Ensemble Model

With the continuous improvement of China’s transportation network,the transportation of freight is increasingly frequent.As a specific indicator reflecting the achievements of transportation production,the volume of freight can embody the quantity of transportation services for national economic and social development.It is not only related to the normal operation of the social and economic system,but also related to the security and stability of the country.Therefore,the accurate forecasting of freight volume is greatly significant for the normal operation and management of transportation companies.Due to natural disasters,industrial level,political events and many other factors,the main characteristics of freight volume include the emergence,high volatility,non-stationary and non-linear.The existing research results of the hybrid forecasting of freight volume are quite abundant,but they are far from perfect.With a view to obtaining more accurate forecasting of freight volume,the commonly used methods are to decompose the original data into several modal branches with different characteristics.However,these branchees are still unstable.For the sake of solving the problem,a second decomposition-ensemble method is proposed.In this paper,the original data of freight volume are decomposed by a proper decomposition method to obtain several modes with different characteristics,which initially reduces the difficulty of prediction.Then,the complexities of the modal branches are measured by sample entropy.According to the level of complexity,these modes are divided into different modal groups.There are high-frequency and complex in the most complex mode group,which have a tremendous influence on the forecasting results.In view of this situation,it is necessary for the most complex modal group to conduct the secondary decomposition.The further decomposition can not only fully extract the effective information hidden in the high-frequency data,but also additionally reduce the complexity of the data.Furthermore,the inputs of neural network are utilized by empirical value.This will result in too many or too few inputs.In order to avoid the interference of invalid information or the issue of insufficient information,feature selection is conducted to design the input form of neural network skillfully.Then,by means of the complexity of different modes,different forecasting methods are used pertinently.Next,the forecasting results of each branch are added to get the ultimate forecasting results.Finally,the stability of the proposed model is evaluated by the Diebold-Mariano(DM)test from a statistical perspective.Considering the influence of different span and training set on the forecasting results,the training set of different ratio and multi-step forecasting are proposed to forecast freight volume.To verify the rationality and practicability of the proposed model,the freight volume data of Shanghai port,Shenzhen port,Guangzhou port,Beijing airport,Shenzhen airport and Shanghai airport are used as research objects.The experimental results show that the proposed model has the high accuracy and strong stability,compared with the benchmark models.Therefore,the proposed model,secondary decomposition-ensemble,can effectively forecast the unstable and complex data of freight volume.