Research On The Forecast Of The Short-term Number Of Cargo Distributing Vehicles In Bulk Cargo Port

Port-highway transportation is one of the main methods of port bulk cargo port transportation in China.In order to ensure the efficiency of Port-highway transportation,ports usually plan resource allocation for vehicle-cargo-distributing operations in advance.However,the plans are manually formulated,leading to a large gap with actual demand,and problems such as traffic congestion during the peak period.This paper provides a reference for the scientific formulation of the plan by predicting the short-term number of cargo distributing vehicles as follows:(1)Analysis of factors affecting the short-term number of cargo distributing vehicles at the bulk cargo port.Through literature review,expert interviews and business analysis methods,identifying influencing factors and feature selecting based on random forest algorithm,mining the key features for subsequent modeling.(2)Construction and verification of the prediction model for the short-term number of cargo distributing vehicles at the bulk cargo port.Introduce the above feature selection results,establish and evaluate a short-term prediction model based on SVM;establish and evaluate a short-term prediction model based on GRU according to the time series of the short-term number of cargo distributing vehicles.(3)Construction and application of short-term forecast combination model.In order to strengthen the performance of the model under complex problems,the SVM-GRU combination model was constructed and verified with Guangzhou Port Group as an example.The SVM-GRU model was applied to the business system to provide a reference for resource planning.This paper proposes an SVM-GRU combined model for the prediction of the short-term number of cargo distributing vehicles in bulk cargo ports,with a MAPE of7.0031.In the analysis of influencing factors,it was demonstrated that the ships’ pre-reported information and the unloading volume of the warehouse have a greater impact on the cargo-distributing operations.Besides,the concept of spatial saturation is introduced.Adding spatial saturation into the model,the prediction accuracy is improved by a reduction of 3.6878 of MSE.The combination model proposed in this paper has certain reference value for the formulation of the resource allocation plan.