A Survey on Forecasting Giant Frequency in FoRex Exchange Rate by Execution of Linear Models

Abstract

The currency exchange rate is one of the most important determinants of a country's relative level of economic health, to predict the direction of price movement for currency is the most important factor. In existing paper ANNs, GFNN and SVR were used to predict the volatility in exchange rate. NN models used dynamic training sets using sliding window technique and a set of threshold functions connected to each other with adaptive weights to get latest data for predicting the model. Whereas the integration of GF with NN is that the GA provides initial weights for FNN by this it not only decrease the training time but also avoid the local minimum. SVR shown better performance when compared to NN based models, in SVR with linear kernels it used static training set of historical data where the parameters are real and observable data values.