A general heuristic for the design of neural networks in financial domains is that the more knowledge that is available to the neural network for forming its model, the better the ultimate performance of the neural network [11, 38]. with a minimum of two years of training data a nominal starting point. Box et al. [5] indicate that time series models improve as more data is incorporated into the modeling process. Technical analysis estimates for various financial time series indicate that historical information is required for anywhere from one year 1171 to six years 16]. Others [30] have indicated that currency exchange rates have a long-term memory, which implies that larger quantities (periods of time) of data will produce more comprehensive models and produce better generalization.