Short-term Prediction Of Bulk Cargo Port Throughput Based On Time Series Analysis

The port is important for domestic and foreign logistics,and the increasingly competitive environment has prompted it transform into a smart one.Throughput is important for measuring the production of enterprises.Long-term throughput forecast helps companies develop investment plans and strategies.The short-term forecast of throughput helps enterprises to arrange production activities and adjust the business rhythm of enterprises,which is of great significance to the development of smart ports.We use deep learning technology to process the monthly throughput data of bulk cargo ports,and research the short-term prediction of throughput from the perspective of time series analysis,which mainly includes the following aspects:(1)This paper analyzes the characteristics of the time series of bulk cargo port throughput and verifies the composite trend of linear and nonlinear.Combined with the business characteristics of bulk cargo transportation,the long-term,seasonal,and volatility characteristics are analyzed,and the influencing factors of the above features are extracted for subsequent modeling.(2)Construction and verification of ARIMA linear throughput prediction model and LSTM nonlinear throughput prediction model.The optimal parameters of the model are determined by the grid search method.Multi-feature influencing factors are added to the LSTM model to further improve prediction accuracy.(3)Construction and application of throughput prediction combination model.In order to avoid the shortage of the single prediction model in the composite problem,the linear model is combined with the nonlinear model by weight distribution.We add the combined model to the Guangzhou Port Production Business System to assist the company in making decisions.This paper verifies the applicability and advantages of the linear model ARIMA and the nonlinear LSTM model.The average relative error of the combined model is 0.096,which is much lower than the 0.1819 of the third exponential smoothing method.In addition,the RMSE,MAE and error levels are selected to evaluate the prediction effect,the superiority of the combined model is demonstrated.The throughput prediction combination model has certain reference significance for port enterprises.