The Forecasting Of Exchange Rate Using Support Vector Machine

In the process of economic globalization, as the relative price of two currencies, the exchange rate has played a role as the bridge to maintain economic exchanges between the two or more States. And this effect is strengthened day by day. Exchange rate, which is contacted with various of macroeconomic factors, microeconomic factors and affecting the national economy's internal and external balance, is one of the core economic variables in the open economy. Therefore, for the state or other economies, forecasting the exchange rate has important practical significance. The exchange rate forecasting is also with practical value for the enterprises, other organizations and individuals.In view of this, this paper used the exchange rate theory and support vector machine theory as the main theoretical basis, set up a index system for the forecasting of exchange rate, constructed a model using principal component analysis and support vector machines to forecast the exchange rate. In chapter 2, introduced the background, assumptions and conclusions of the five kind of exchange rate theory from which sort out the factors that influence the exchange rate. Outlined the basic principles of support vector machines, described the basic principle of the principal component analysis and data processing briefly. In chapter 3, according to certain principles of indicators selection identified indicators. Then established the index system to forecast the exchange rate. Combined SVM and PCA, and then defined the the structure and processes of the composite model. In order to verify the effect of the exchange rate forecasting model, implemented the model in the software platform, picked out the exchange rate of RMB against the U.S. dollar to study, then the real exchange rate data were forecasted in chapter 4. The result was compared with the support vector machine model and GA-BP model. Comparative results showed that the PCA-SVM model, the MAPE of which is 0.38%, was more accurate. The forecasting results illustrated the ideas of constructing index system in this paper is correct. The reslut also proved the effect of principal component analysis is not only simplify the input of SVM model, but also can improve the forecasting accuracy.