LAUSR

To improve the RMB exchange rate prediction and risk measurement, the RMB exchange rate prediction model is constructed based on deep learning approaches. Value at risk (VaR) risk measurement related data are used, and this model is combined with the autoregressive moving average model-generalized autoregressive conditional heteroskedasticity (ARMA-GARCH) model to fabricate an integrated VaR risk measurement model. The effectiveness of the proposed model is verified on specific example data. The results show that the proposed deep learning RMB exchange rate prediction model has better performance than traditional exchange rate prediction models in predicting exchange rates in different international foreign exchange markets, with accuracy of 74.92%. ARMA-GARCH risk prediction model has good measurement performance for the market, and its accuracy is significantly higher than that of the traditional measurement model. The deep confidence network model has stable performance and ideal forecasting effects both in the forecast of exchange rate fluctuations and in risk measurement. In short, this research can improve China’s research on exchange rate fluctuations and effectively strengthens the ability of forecasting and risk assessment of the foreign exchange market.