Comparison of neural networks for foreign exchange rate prediction

Abstract

The paper incorporates a comparison between long short term memory (LSTM), simple recurrent neural networks (SRNN), and gated recurrent unit (GRU) to predict currency exchange rate for 22 countries-currencies against United States Dollar(USD) simultaneously. The models predict foreign currency exchange rates for 30 consecutive days by taking the last 365 days of data as input. The models work with the same number of neural network layers, input, targeted output, optimizer, and learning rate.