Research on logistics volume prediction model based on ARIMA-LSTM

Abstract

At present, the line volume prediction problem in the cargo transportation problem is very important, if the line volume prediction in the parcel emergency transportation and structure optimization problem in the e-commerce logistics network is not handled well, it may lead to unreasonable allocation of logistics resources, logistics congestion, parcel delays, and even loss. Meanwhile, in order to increase the transportation efficiency of logistics network, this paper proposes an ARIMA-LSTM prediction model. Through data analysis and processing, it is found that the data is characterized by time series data. For this reason, this paper firstly uses the ARIMA model to predict the cargo volume, and finds that it can only predict the linear part of the cargo volume well. Secondly, the LSTM model is used for prediction, and contrary to the ARIMA model, it is found that the LSTM model can better capture the information of the nonlinear part of the cargo volume; finally, this paper combines them to construct the ARIMA-LSTM model, and finally compares the prediction effect of these three models by using the MAE, MSE, and R2, and it is found that the prediction effect of the ARIMA-LSTM model is the best, and the MAE is 0.5. The ARIMA-LSTM model was found to be the best. The MAE is 0.00042, the MSE is 3.08e-08, and the R2 is as high as 0.89.