Predictive Analysis Of Express Business Volume Based On Time Attention Mechanism And LSTM Neural Network

The express delivery industry plays a vital role in the entire service industry,so how to accurately estimate its business volume has become a hot topic in today’s research.The accurate prediction of express delivery business is of great significance to the rational planning of logistics layout,the construction of logistics facilities,the adjustment of economic structure,and the optimization of enterprise production planning and resource allocation.Using the Python 3.9 programming language,this paper conducts an in-depth prediction analysis of China’s monthly express business volume from January2010 to September 2022,and puts forward a series of effective prediction suggestions in order to obtain more accurate results in subsequent research.The main research contents are as follows:Based on the compilation and research of relevant literature,this paper analyzes various factors affecting express business volume,and then establishes a preliminary express business volume prediction index system.The time series chart and monthly distribution map of the monthly data of express delivery business volume are plotted by Python programming language,and the results indicate that China’s monthly express business volume shows the characteristics of year-by-year growth,random fluctuation and seasonality.Then,the random forest method was used to supplement the missing data values,and the Spearman grade correlation coefficient was used to analyze the correlation between each factor and express delivery business volume,and finally an index system for predicting express delivery business volume was created,which included seven indicators: postal business volume,cargo turnover,freight volume,road freight volume,Internet broadband users,total import and export volume and total retail sales of consumer goods.Postal business volume is used to measure the development of the postal industry,cargo turnover,cargo volume and road freight volume are used to measure the level of transportation,the number of Internet broadband users can explain the level of information technology development,the total import and export volume and the total retail sales of social consumer goods represent the level of domestic and foreign trade.Four models are established: the backpropagation neural network model(BP),two recurrent neural network models(RNN and GRU),and the long short-term memory recurrent neural network model(TATT-LSTM)with temporal attention mechanism,and after adjusting the parameters,they were used to predict China’s monthly express business volume data,and the prediction effect of the comparison model was evaluated through error evaluation and goodness-of-fit.The results show that the four prediction errors(MSE,RMSE,MAE and MAPE)of the TATT-LSTM neural network model are the smallest,followed by the GRU neural network model.The prediction accuracy of the TATT-LSTM neural network model is0.81,which further shows that the LSTM model with time attention mechanism can predict the monthly express business volume in China better,and then provide reference for the government,postal administration and express delivery related enterprises to formulate development strategies.