Prediction Of Express Business Volume Based On Stacking Algorithm

Courier industry is an important part of service industry,and forecasting express business volume is an important topic in the research field of courier industry.Therefore,in this paper,the data related to the monthly express business volume in China from January 2008 to May 2021 are predicted and analyzed in Python 3.7 environment.The main research work is as follows.1.The Spearman rank correlation coefficient and distance correlation coefficient are used to analyze the correlation between each influencing factor and express business volume,and finally the index system for express business volume forecasting is constructed,which contains six indicators of total retail sales of social consumer goods,the number of Internet broadband subscribers,civil aviation cargo volume,gasoline production,power generation and postal business volume.2.A single time series analysis is done for the express business volume series,and the experimental results: the average absolute percentage errors(MAPE)of the model,HoltWinters seasonal addition and multiplication models are 8.646%,36.6% and 38.8%,respectively,and the model has the smallest percentage of the other three errors,and the highest prediction accuracy is 0.88.3.Do predictive analysis based on Stacking algorithm for data related to express business volume.By using Bayesian optimization,sparrow search algorithm and Circle mapping improved sparrow search algorithm to optimize the hyperparameters of seven machine learning regression algorithms,the final result of the cross-validation method is that the base regressor is GBDT algorithm,RFR algorithm,SVR algorithm and metaregressor is Ridge regression algorithm based on the Stacking algorithm model has the best prediction effect,the prediction accuracy is The prediction accuracy is 0.945 and the MAPE is 8.587%,comparing with the model analysis,the prediction accuracy of the former is 6.5% higher and the MAPE is 0.059% lower than that of the latter,indicating that the prediction model based on the Stacking algorithm introduced has good applicability for the prediction analysis of China's monthly express business volume data.The quantitative analysis of the characteristic variables by SHAP model concludes that the number of Internet broadband access users(informatization factor)is the key influencing factor of express business volume.