Predicting the Total Export-import Volume of China's Economic Trade through Machine Learning

Abstract

By forecasting the total export-import volume of China's economic trade, valuable guidance can be provided for the formulation of relevant policies, thus holding significant practical importance. This paper first analyzed the current situation of the import and export of China's economic trade. Subsequently, key indicators such as gross domestic product (GDP) and producer price index (PPI) were selected. After eliminating irrelevant indicators through correlation coefficient calculation, seven remaining indicators were employed for research. Building upon the machine learning algorithm-support vector machine (SVM), an improved sparrow search algorithm (ISSA) was developed to optimize SVM parameters, forming the ISSA-SVM prediction method. Experimental validation was conducted using data from 2003 to 2022. The results revealed that the average time consumed by ISSA-SVM for forecasting was 1.742401 s, displaying a minimal difference compared to the SVM method. In terms of total volume prediction, the ISSA-SVM approach achieved a mean absolute percentage error of 0.02% and a root-mean-square error of 1068.25, surpassing logistic regression, back-propagation neural network (BPNN), and other methods. These outcomes verify the reliability of the ISSA-SVM method in total volume prediction, showcasing its practical applicability.

Keywords

• machine learning
• china’s economy
• total export-import volume
• trade
• forecasting