Forecasting of Currency Exchange Rates Using ARIMA, ABC with DNN

Abstract

The foreign exchange market is one of the largest markets in the world. In this market, the selling and buying of another currency takes place, which is crucial for currency trading on the international market. In this paper, an Instant Prediction System (IPS) is proposed to predict Foreign Exchange Currency rate (FOREX) based on deep learning with optimization and Autoregressive integrated moving average (ARIMA) model. ARIMA is a popular linear time series model used for forecasting for the last couple of years. The present research, with Deep Neural network (DNN) and Artificial Bee Colony (ABC) as an Optimization Algorithm is a promising and highly accurate alternative to the traditional linear methods. By considering the advantages of ARIMA, DNN with ABC model in a linear and non linear modeling a hybrid model is designed. Experimental results are computed with the real time dataset obtained from FOREX site http://fx.sauder.ubc.ca/data.htmlfrom the year (01/01/2015-31/12/2018) and the indices such as Mean average Precision (MAP), Mean Absolute Percentage (MAPE) Error and accuracy are evaluated. The average prediction accuracy examined for the proposed IPS system is 98.22%.
Keywords: Foreign Exchange Currency rate, ARIMA, ABC, DNN.