A Method Study Of Forecasting The Traffic Volume Of The Mechanical Refrigerator Wagons Of The Cold Logistic Chain In China Railway Special Cargo Logistics Company

With the rapid development of the social economy and the gradual improvement of people’s living standards,much attention has been drawn to food safety,especially railway cold-chain logistics.Within the cold-chain logistics industry in China,railway mechanical refrigerator cars(RMRCs)play an essential role in transporting China Railway Special Cargo Services CO.,LTD(CRSCS).Compared with road transport,RMRCs have the advantages of the temperature control and high-quality transport;However,the disadvantage is that the transportation volume is large,the supply of goods needs to be assembled,the transportation time is long,the timeliness is poor,and there are short barge loading and unloading operations at both ends,and the cost of entire transportation is expensive.Therefore,it is vital to build a model for forecasting the daily number of RMRCs that will greatly improve the efficiency of the utilization of RMRCs.Moreover,the model can help enterprises to formulate a reasonable operation plan for RMRCs,and strengthen the management level on transportation projects of RMRCs.By analyzing the previous literature on forecasting railway cold chain logistics,the selection of indicators and methods in domestic and foreign are gradually completed,and the prediction accuracy is also continuously improved.It is hard for grey forecasting to predict the long-term demand,the exponential smoothing method has low forecasting accuracy,and the time series method has strong fluctuations.Therefore,this paper uses the Gated Recurrent Unit(GRU)algorithm to forecast the number of used mechanical refrigerator cars in cold-chain logistics.The RMRCs of CRSCS was used as the research object.This paper analyzes the overall characteristics of the traffic volume of RMRCs,and reveals the RMRCs traffic volumes of different Railway Bureau group companies,including the characteristics of arrival traffic volume and departure traffic volume.Then,the traffic volume characteristics of RMRCs for different categories of products were uncovered.Finally,the paper establishes a scientific and accurate model to predict the traffic volume of RMRCs by using the daily traffic volume of RMRCs of CRSCS data from January 2017 to October 2020.When analyzing the traffic volume of RMRCs by using the data in this time period,this paper finds that the data has a strong periodic rule.Therefore,the traffic volume of RMRCs can be treated as a time series,and the traffic volume of the specified day is greatly affected by the traffic volume of previous days.Compared with the multiple linear regression(MLP),support vector regression(SVR),random forest(RF),artificial neural network(ANN),and GRU algorithms,it is finally determined to establish the traffic volume prediction model of RMRCs by GRU algorithm.By analyzing the residual Q-Q diagram and residual distribution histogram of the established model based on the GRU algorithm in the test dataset,it can be seen that the residual of the prediction model results in the test dataset does not contain unprocessed information,the overall distribution of the residual conforms to the normal distribution,and the prediction result of the model established by GRU algorithm is relatively reliable.The GRU model has a good prediction performance on the different number of RMRCs and has a good prediction accuracy when predicting the traffic volume of RMRCs calculated by month.