Research On Exchange Rate Forecast Based On EMD And Event Impact

With the deepening of economic globalization,the fluctuation of a country's economy will seriously affect the trend of global economic development.Foreign exchange market plays an important role in the economic exchanges of various countries.Since the implementation of China's exchange rate reform policy in July2005,the RMB exchange rate formation mechanism has become increasingly market-oriented,and the market factors affecting exchange rate fluctuations have gradually increased,which makes it more and more difficult to predict the trend of exchange rate.Effective forecasting of exchange rate changes can not only enhance the effectiveness of the state's supervision and management of foreign exchange market,but also play an important role in maintaining international financial stability,international investment of enterprises and risk management of individual foreign exchange investment.Exchange rate series is a non-stationary and non-linear complex series.Traditional linear and non-linear forecasting methods can only forecast the exchange rate series from a single scale,resulting in poor long-term forecasting results.Empirical mode decomposition(EMD),as a commonly used multi-scale decomposition method,has good time-frequency characteristics,and can decompose time series on different time-frequency scales,so as to analyze its intrinsic characteristics.In addition,the simple structured time series forecasting model ignores the impact of special events in the foreign exchange market on the exchange rate,which results in the decline of the forecasting accuracy of the model in the volatility period.To solve the above problems,this paper proposes an exchange rate forecasting model based on EMD and event effects.The first task of this paper is to sort out the existing exchange rate forecasting methods in detail.Through analysis and comparison,EMD decomposition-combination forecasting model is proposed to realize the forecasting of exchange rate time series data.After studying the shortcomings of the traditional EMD method,this paper proposes to improve the decomposition effect of exchange rate series by using Monte Carlo wavelet threshold denoising method;reconstructs the component series by using an adaptive center DTW algorithm,and obtains the sequence of high-frequency,low-frequency and trend items of exchange rate series;finally,according to the different characteristics of the reconstructed sequence,different predictions are selected.The model realizes combination forecasting.On this basis,considering the impact of special events in the foreign exchange market on the exchange rate,the event-based exchange rate forecasting model is constructed.Firstly,by defining and classifying special events in the foreign exchange market,the event ontology library is constructed.Secondly,the method of calculating the impact degree based on event ontology library is introduced.Finally,by calculating the impact degree of events,the prediction of the impact of market events on the exchange rate is realized.Finally,a support vector regression model is proposed to integrate the two forecasting results,and a complete exchange rate forecasting model based on EMD and event impact is constructed.In this paper,a large number of comparative experiments verify the validity of the research content.The experimental results show that adding the influence of special events to the structured exchange rate time series prediction can effectively improve the prediction accuracy of the model.