Prediction and Analysis of Import and Export Volume Based on LSTM Recurrent Neural Network and Kruskal-Wallis H-Test

In the context of globalization, it is important to accurately forecast the total amount of imports and exports. This paper first analyses the growth of China's total import and export and M2 supply from 2000 to 2024. The results show that the total amount of import and export shows a linear growth, with the regression equation <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$y=2 \times 10^{6} x-2 \times 10^{9}, R^{2}=0.9249$</tex>. While the M2 supply shows exponential growth, with the regression equation <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$y=0.1319 e^{0.0115 x}, R^{2}=0.9832$</tex>, which indicates that the M2 supply grows at the average annual rate of 1.15 %. This paper further establishes a forecasting model of total import and export based on an LSTM-fully connected neural network, corrects the LSTM output with the M2 data input into the fully connected layer of the previous period, and conducts a lagging experiment, which shows that the lagging will lead to a decrease in the total amount of import and export. Finally, the significance of the impact of the money supply was analyzed using the Kruskal-Wallis H-test, and the results proved that its impact was significant. The research in this paper provides a reference for the future forecast of total import and export amount and related decision-making.