Research On RMB Exchange Rate Forecasting Based On Comprehensive Model

With the multiple reforms of the exchange rate system,the degree of influence of politics,marketization and sudden information on the RMB exchange rate has continued to expand.This leads to the more obvious randomness and uncertainty of exchange rate fluctuations,which makes it more difficult to accurately analyze the characteristics of exchange rate fluctuations and predict the future trend of exchange rate fluctuations.Aiming at the above problems,this article constructs a comprehens ive CSERAL model based on a decomposition-reconstruction-integration based Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)method,a sample entropy reconstruction(SER)method,an Autoregressive Integrated Moving Average model(ARIMA),and a Long Short-Term Memory Neural Network model(LSTM).This model is applied to the exchange rate data processing of USD/CNY,EUR/CNY and JPY/CNY.The specific research contents and work are as follows:(1)In terms of data decomposition,the Empirical Mode Decomposition method can decompose complex non-linear data into simple sequences with obvious fluctua t io n characteristics,and build a model on this basis for prediction,which can effective ly improve the model’s ability to predict data.Aiming at the complexity and random uncertainty of the exchange rate fluctuation curve,a CEEMDAN decompositio n method is constructed to decompose it,discarding the useless redundant informat io n,obtaining a low-dimensional sequence with obvious fluctuation characterist ics,reducing the complexity of subsequent data processing.This lays the foundation for subsequent data processing.(2)In the aspect of data reconstruction,a reconstruction method based on SER is constructed to solve the problem that too many decomposition components can easily lead to the accumulation of errors in the model system.It reduces the dimensions of the components and obtains reconstructed components that explain their economic significance and have a driving effect on exchange rate data.From this,low-frequenc y sequences representing major events,economic cycle changes,trend item sequences representing long-term trends,and high-frequency sequences representing short-term random factors can be obtained.This method effectively integrates the influence degree information of the three sequences,and deeply focuses on multi-dimensional features such as RMB exchange rate changes and fluctuations,thereby reducing the systemat ic error of the model and laying a foundation for improving the subsequent prediction model.(3)In terms of data prediction,a prediction model based on ARIMA and LSTM is constructed to solve the problem that a single model cannot ful y exploit the informa t io n of reconstructed sequence data.On the basis of the CEEMDAN decomposition method and the sample entropy reconstruction method,the ARIMA model is used to extract the features of the high-frequency sequences,the LSTM model is used to extract the features of the low-frequency sequences and the trend item sequences,and the feature information obtained by the two is fused to obtain the final prediction value.According to the statistical evaluation results,the comprehensive model proposed in this paper achieves better prediction accuracy,and can give full play to the prediction advantages of different models,which greatly improves the prediction ability of the comprehens ive model,provide a more reasonable reference for the government to formulate macro policies and individual investment.