In this study, by using back-propagation learning of the neural network, the exchange rate fluctuations in the future is predicted. In order to predict the exchange rate, optimal combination of intermediate layer and parameters was investigated.
then, the network predicted the exchange in 2014. As a result of learning, it was not able to sufficiently predict by small error This means that prediction failed.
Therefore, it predicted again by changing the supervisory signals. As a result, it has been able to show the trend of changes in exchange,but can not be more predictive. The future task is to improve the way to learn the detailed information for more accurate prediction.