A Study of the Dollar Index Forecast Based on the SVM Theory
|Keywords||dollar index support vector machine factor analysis|
Since QE, the USD goes unclear, while the foreign reserve of China rises every year with a rise of foreign trades and foreign investment So the trend of dollar affected our economy directly which makes it meaningful to study the method of dollar index forecasting.This article mainly uses the support vector machine to study the dollar index forecasting. Based on the exchange rate theory and support vector machine, this article picks indexes that effect the dollar index to build the index system, then uses the factor analysis to simplify the index system. Finally, we used the parameter optimized SVM to forecast the dollar index.In the empirical part, this article uses PSO method, genetic algorithm and grid method to optimize the parameter of the SVM model. Then we compare the result with each other. Totally, we got the expected result. Use the foreign exchange theory, we build the index system successfully, and the factor method simplified the input variables. Finally, with the three different SVM models, this article ran out with a relatively accurate result. Among them the result by genetic algorithm is the best, which carries out an average error rate of0.393556%and a minimum error rate of0.004165%. Finally this article compared the result of FAC-SVM model with BP neural network and normal SVM model, and proved the advantage of the model and the effectiveness of the index systems.