LAUSR

In this paper, we introduce a model based on Convolutional Neural Network for forecasting foreign exchange rates. Additionally, a method of transforming exchange rates data from 1D structure to 2D structure is proposed. The transaction of the foreign exchange market has periodic characteristics, however, due to the technical limitations, these characteristics cannot be utilized by existing time series forecasting models. In this paper, we propose a model which can process 2D structure exchange rates data and put these characteristics to good use. Exchange rates Euro against US dollar, US dollar against Japanese yen and British Pound Sterling against US dollar are researched in this paper. Our experimental results show that, when compared with Artificial Neural Network, Support Vector Regression and Gated Recurrent Unit, the proposed model can effectively improve the accuracy of long-term forecasting.