Prediction on China's merchandise exports based on BP neural network associated with sensitivity analysis

摘要

In this paper, six impact indicators, including exchange rate of RMB, gross domestic product (GDP), consumer price index (CPI), foreign direct investment (FDI), investment in fixed assets (FAI), and research and development (R&D), which seriously affect China's merchandise exports, were employed as the input variables, and China's total exports as output variable to establish a BP neural network model, in order to predict the country's total exports. To test the accuracy of the BP model, the parameter Nash-Sutcliffe model efficiency coefficient (NSC) is introduced and found that this model has high accuracy (NSC >0.9957). In addition, indicators GDP, FDI, and R&D were found to influence China's total merchandise exports significantly through the sensitivity analysis. The method above along with policy analysis could systematically analyze and predict China's exports, providing a theoretical basis for relevant decision-making departments.